CN114238772B

CN114238772B - Content-adaptive and aware web map intelligent recommendation system

Info

Publication number: CN114238772B
Application number: CN202111603724.1A
Authority: CN
Inventors: 韩效遥
Original assignee: Individual
Current assignee: Individual
Priority date: 2021-12-24
Filing date: 2021-12-24
Publication date: 2024-11-12
Anticipated expiration: 2041-12-24
Also published as: CN114238772A

Abstract

The application integrates image content-based clustering and image recommendation into network map service retrieval and recommendation, greatly improves map service search experience in the prior art, improves service discovery efficiency, provides a whole set of technical processes of data collection processing, thumbnail generation, image clustering and platform construction, realizes network map service automatic sampling based on information quantity and approximation degree, color and texture-driven image clustering map recommendation, designs a map service retrieval recommendation platform for realizing a Web end on the basis, has accurate grasp and identification of map content, can conduct network map intelligent recommendation based on content adaptive perception, develops a map service retrieval platform based on keyword search and image content recommendation, has high matching degree of retrieval map content and actual requirements, verifies rationality and effectiveness of the method, and has higher practical value.

Description

Network map intelligent recommendation system with self-adaptive sensing content

Technical Field

The application relates to an intelligent network map recommendation system, in particular to an intelligent network map recommendation system with self-adaptive sensing of content, and belongs to the technical field of intelligent network map content recommendation.

Background

The network map service is a novel map organization mode in the mobile Internet era, integrates the characteristics of multi-scale coupling, high-level interoperability and the like on the basis of basic map visualization, enters a big data era, and becomes the optimal knowledge expression and display service in a lower resource demand, visual visualization form and seamless cross-platform connection. In recent years, the number and quality of map services are continuously improved, and the demands of the public for map services are gradually improved, so that how to effectively manage the map services and quickly match the demands of users becomes a research and development and application hotspot.

Currently, a large amount of geographic information data, geographic information networks, geographic information portals, geographic information documents and the like are scattered on heterogeneous network nodes, and form massive network geographic information resources. Meanwhile, the units and individuals can carry out auxiliary decision making, application management, auxiliary scientific research and the like by acquiring various geographic information resources. Among them, the network map service represented by OGC WMS is widely applied in various industries, and how to quickly search for a map service meeting the requirements is a current urgent problem to be solved.

Compared to traditional keyword text-based searches, image content-based searches have a better visual experience and more accurate search results. Hundred degrees, google, etc., have proposed a search function based on image content. If the two internet ideas are introduced into the map service, a map service recommendation function based on image content is developed, a keyword-based search function can be effectively assisted, the map service search experience in the prior art is improved, and the service discovery efficiency is improved.

The content-aware map recommendation system is an important and effective supplement to the text-based service search in the prior art, and the image content-based recommendation function makes up for a single geographic information service discovery mode based on keyword search in the prior art, and has important effects and great application value in enhancing user experience and improving map service discovery efficiency.

Image retrieval is initially mainly to retrieve images through text information, such as shooting dates of images, authors, categories and sizes of images, and the like, and such image retrieval modes are text-based image retrieval technologies. Over ten years of development, image retrieval has evolved to content-based retrieval, primarily retrieving images by their color, texture, shape, etc. features, with the objective of avoiding the use of textual descriptions, but instead retrieving similar images by user-provided query images or user-specified image features based on visual similarity.

Text-based image retrieval is now well established, but has its drawbacks and disadvantages: firstly, labeling texts and keywords of images is completed manually; secondly, the manual annotation has the problems of different observation angles, subjective idea difference and the like, and annotation ambiguity can be generated, namely, the manual annotation is difficult to fully express an image containing a plurality of targets and even covering a plurality of emotion colors. The text-based image retrieval still has the defects, and as market demands and applications are improved deeply, image retrieval applications based on image contents such as hundred-degree image recognition, google image recognition and the like are sequentially presented, and although accuracy is still to be improved, image features can be automatically extracted, so that ambiguity is avoided.

In summary, the network map recommendation system in the prior art has obvious defects, and main defects and design difficulties of the network map recommendation system include:

Firstly, in recent years, the number and quality of map services are continuously improved, the demands of masses on the map services are gradually improved, and the prior art cannot effectively manage the map services and quickly match the demands of users; at present, a large amount of geographic information data, geographic information networks, geographic information portals, geographic information documents and the like are scattered on heterogeneous network nodes, so that massive network geographic information resources are formed, units and individuals can carry out auxiliary decision making, application management, auxiliary scientific research and the like by acquiring various geographic information resources, map services meeting requirements cannot be quickly searched and found in the prior art, and a more intelligent map searching recommendation method based on content is lacked;

Secondly, compared with the traditional keyword text-based search, the image content-based search has better visual experience and more accurate search results, the internet thought is not introduced into the map service in the prior art, and the map service recommendation function based on the image content is lacked, so that the map service search experience in the prior art is poor, and the service discovery efficiency is low; because the content-aware map recommendation is an effective supplement to the text-based service search in the prior art, the image content-based recommendation can make up for the geographic information service discovery mode of the single keyword search in the prior art, has an important effect and great application value in enhancing the user experience and improving the map service discovery efficiency, but the network map intelligent recommendation based on the image content in the prior art is difficult to realize;

Third, prior art text-based map image retrieval is well established, but has its drawbacks and shortcomings: firstly, labeling texts and keywords of images is completed manually; secondly, the manual annotation has the problems of different observation angles, subjective idea difference and the like, and annotation ambiguity can be generated, namely the manual annotation is difficult to fully express an image containing a plurality of targets and even covering some emotion colors, text-based image retrieval still has defects, and the image retrieval application based on image content is sequentially carried out along with the improvement of market requirements and application, but cannot be applied to a map, or the problems of low accuracy, incapability of automatically extracting image features, excessive ambiguity and the like exist;

Fourthly, a plurality of problems exist in the realization of network map recommendation based on content perception, namely, a feasible network map service automatic sampling method based on information quantity and approximation degree is lacked, a map service image sampling method under the constraint of the information quantity and approximation degree is lacked, a plurality of sampling images cannot be generated for each image layer, and an image library cannot be established by using the sampling images; secondly, a feasible image clustering map recommending method is lacked, a method for classifying the image layers based on a clustering method of an image group is lacked according to the characteristic design of a plurality of sampling images of each image layer, and a similar image layer recommending method considering the characteristics of the image group of the image layers is lacked; and thirdly, a feasible network map service retrieval recommendation platform design and development method is lacked, and a high-quality and high-efficiency network map intelligent retrieval recommendation platform is lacked.

Disclosure of Invention

Aiming at the defects of the prior art, the application integrates the clustering and image recommendation based on image content into the retrieval and recommendation of the network map service, greatly improves the experience of map service search in the prior art, improves the service discovery efficiency, provides a whole set of technical processes from data collection processing, thumbnail generation, image clustering and platform construction, realizes the automatic sampling of the network map service based on information quantity and approximation, realizes the image clustering map recommendation based on color and texture, designs a map service retrieval recommendation platform of a Web end on the basis, has accurate grasp and identification of the map content, can carry out the intelligent network map recommendation based on content self-adaptive perception, develops a map service retrieval platform based on keyword search and image content recommendation, has high matching degree between the retrieval map content and actual requirements, verifies the rationality and the effectiveness of the method, and has higher practical value.

In order to achieve the technical effects, the technical scheme adopted by the application is as follows:

the intelligent network map recommending system with self-adaptive content sensing integrates image content based clustering and image recommending into network map service searching and recommending, builds one set of technological process from data collecting process, thumbnail generating process, image clustering and platform, and designs Web terminal map service searching recommending platform based on the technological process, and includes mainly:

First, a network map service automatic sampling method based on information quantity and proximity, comprising: firstly, calculating and evaluating map tile information, and secondly, calculating the approximation degree of the map tiles with cross scales; positioning the rich-content area in the map by adopting an information quantity combined quadtree structure, wherein the information quantity is approximately expressed by the complexity of the image; the approximation degree is calculated through a color quantization unified histogram and is used for screening scaling factors with mutation, and the scaling factors are respectively used for realizing orderly traversal of map services from a plane position dimension and a scaling factor dimension, so that effective data is provided for clustering recommendation based on image content;

second, color and texture driven image cluster map recommendation, comprising: the map service recommendation method comprises the steps of firstly, collaborative representation of map features integrating color and texture features, secondly, clustering of map content images, thirdly, a clustering method based on an image group, and thirdly, recommending map services based on the image group; comprehensively representing image features by adopting color features and texture features, wherein the color features are represented by adopting a color quantization unified histogram, the texture features are represented by adopting a gray level co-occurrence matrix, and the similarity of the inclusion relation and the colors of image groups is considered, and the image group recommendation method reduces the similarity difference caused by image layer positioning by giving different weight values to the different approximation groups, so as to provide a multi-feature representation method for comprehensively representing the colors and the textures, and realize the map service recommendation based on the image groups;

thirdly, building an intelligent retrieval recommendation platform of the network map, which comprises the following steps: firstly, a map intelligent retrieval recommendation platform architecture, secondly, network map data acquisition and processing, thirdly, a map keyword retrieval library, fourthly, map image preprocessing and image retrieval, and thirdly, network map intelligent retrieval recommendation platform building access; a large number of map services are crawled by compiling a web crawler, metadata of the map services are analyzed and stored in a database, a keyword retrieval library is built based on Apache Solr services, keyword retrieval services are issued, thumbnail images are generated for each service and stored, color and texture characteristics of each thumbnail image are calculated and stored in a characteristic library, and clustering results are stored in a category library. When recommending similar images, firstly finding the category of the similar images and then finding the most similar images under the category to recommend the most similar images to the user.

The intelligent network map recommendation system based on the content adaptive perception further comprises a map service preview adaptive sampling method based on the content: in the tile map mode, map service concise expression is defined as screening a plurality of tiles which can represent map service contents from a large number of map tiles, and the map service concise expression is further decomposed into two sub-modules: tile selection in the planar position dimension and tile selection in the scaling dimension;

(1) Tile selection for the planar position dimension: quantizing the representativeness of the map tiles based on a specific scaling, and approximating the information quantity by adopting the complexity factors of the map tiles and taking the complexity factors as quantization indexes of the representativeness of the tiles;

(2) Tile selection for the scaling dimension: the key is to find the tile with mutation, quantitatively evaluate the similarity between the two scale tiles by adopting the image approximation degree, and judge the scaling with obvious change based on the similarity;

based on two quantization indexes of complexity and approximation, two rounds of screening are used for respectively realizing full coverage of plane position dimensions and automatic selection of tiles with multiple scales of scaling dimensions:

The 1 st round of screening, from the minimum scaling, each stage only selects the most complex tile, and uses the geographical range of the tile as constraint to process the next scaling, the first round of screening filters out most tiles;

round 2 screening: comparing the approximation degree of tiles between the scaling ratios, further removing redundancy, defining a sampling set as R, and specifically comprising the following steps:

Step one: starting from level L ₀, adding the only map tile in level L ₀ to R and using the map tile as a seed tile;

Step two: requesting 4 map tiles of the same geographic range as the seed tiles at a higher level, and taking the tile Ringo with the most abundant information content as a new seed tile;

Step three: splicing 4 tiles to obtain Rpvda, if the approximation degree of Rpvda and the seed tiles is larger than a critical value, adding Rpvda to R, otherwise, not adding;

repeating the second step and the third step until all the scaling ratios are traversed, and obtaining a final sampling graph set;

a plurality of variables are flexibly adjusted: the number of tiles of adjacent scaling does not necessarily meet the 4-time relation, the number of tiles spliced is affected, the first r (r is more than or equal to 1) tiles with the most abundant information are used as seed tiles, and the probability of obtaining representative tiles is increased.

The content self-adaptive perception network map intelligent recommendation system further comprises the following steps of calculating and evaluating map tile information: the method comprises the steps of 1 st-round screening based on map information quantity, selecting map tiles with the most abundant information quantity under each scaling, using a geographical space range covered by the seed tiles as constraint heuristic type traversing next-stage scaling, and ensuring that the seed tiles of other stages except L ₀ and L ₁ are not guaranteed to have the maximum information quantity, but obtaining higher hit rate with lower calculation cost, wherein the number of map tiles which participate in comparison at each stage after rough selection is not more than 4;

In the aspect of calculation and evaluation of map information, the data quantity of the map tiles after compression is used as a representation index of the information quantity of the map tiles, and the data quantity of the map tiles under the PNG format is used as an information quantity index.

The network map intelligent recommendation system with the content self-adaptive perception further calculates the approximation degree of the map tiles with the cross scale: comparing the approximations of map tiles between two adjacent scaling scales, and judging mutation or slow change according to the approximations, wherein only the map tiles with obvious changes are reserved; analyzing the similarity degree of map tiles from the color features, and quantitatively expressing the color features of the map by adopting a color quantitative unified histogram, and measuring the similarity of the map tiles based on cosine distance weights on the basis;

the application combines map expression characteristics to carry out two-point improvement on the histogram generation method:

Improvement 1, for color quantization: firstly, carrying out color quantization on map tiles, firstly splicing two tiles to be compared together, uniformly managing all colors of the two tiles, and obtaining a remarkable and unique color set through quantization;

improvement 2, calculating the frequency of occurrence of color: firstly, dividing a map into a plurality of continuous color blocks, and calculating the occurrence frequency of corresponding colors by using the circumferences of the color blocks, wherein the dividing standard of the continuous color blocks is that pixels with the same color are adjacent in a four-neighborhood, and a molar neighborhood boundary tracking method is adopted in circumference calculation;

Under RGB color space, color quantization adopts a minimum variance method, the number of reserved colors is specified, the occurrence frequency of each color in two tiles is counted based on quantized color sets, a uniform histogram of color quantization is obtained, and finally the similarity degree between the histograms is calculated by adopting cosine distance; the color weight value with the largest duty ratio in the histogram is set to 0, and the influence on the approximation degree is restrained.

The intelligent recommendation system for the network map with the self-adaptive sensing content further comprises the step of recommending the image cluster map driven by colors and textures: classifying map images according to image features, clustering the map layers to narrow a similar map layer searching range before recommending the map layers, respectively using a color quantization unified histogram and a gray level co-occurrence matrix to represent the color and texture features of the images, and clustering and recommending image groups aiming at network map service features on the basis of generating 3 sampling images for each map layer;

Map feature collaborative representation of integrated color and texture features: for a color map, the color features and texture features of the color map are respectively marked as G _s and G _r, for two given images D and E, the color feature difference between the two images is marked as A _s, the texture feature difference is marked as A _r,k₁ and k ₂, and the weight values of the distances between the two features are respectively marked, so that the comprehensive distance between the D and E is:

A(D，E)＝k₁*A_s(GS_D,GS_E)+k2*Ar(Gr_D,Gr_E)

This is used as a similarity comparison method between the map image and the search.

The network map intelligent recommendation system with the content self-adaptive perception comprises the following steps of: all objects in the map data set are used as cluster centers at the beginning of clustering, each cluster center is updated through a message transmission mode in each iteration, one cluster center is selected as the center of the object which fails in competition after each iteration, and a plurality of high-quality clustering results are finally obtained through continuous iteration and selection;

Clustering is carried out according to the approximation degree among M map data points, the approximation degree forms an M multiplied by M similarity matrix, all map data points are used as potential clustering centers Z, the larger the value C (w, w) on the diagonal line of the matrix C is, the larger the probability that the point w becomes the clustering center is, the value of the point w is the reference degree q, a transmission mechanism mainly comprises an attraction factor and a attribution factor, the attraction factor and the attribution factor matrix are updated, attenuation coefficients k are introduced for avoiding oscillation of results, and when information is updated, each information is set to be k times of the last iteration updated value plus 1-k times of the updated value of the information, wherein the attenuation coefficient k is a real number between 0 and 1. I.e. the iteration value of t (j, w) and d (j, w) at the (r+1) th time:

d_r+1(j,w)←(1-k)d_r+1(j,w)+kd_r(j，w)

the specific steps of the map content image clustering are as follows:

step 1: initializing parameters and reading map content data;

Step 2: calculating an approximation matrix, solving a median value of the approximation matrix and giving the median value to the parameter illuminance q;

Step 3: updating the attraction factor matrix and updating the attribution factor matrix;

Step 4: judging whether the maximum iteration times are reached or the termination condition is met, if so, jumping to the step 2, otherwise, performing the step 5;

step 5: and obtaining a final cluster center Z, and dividing each data object into corresponding class clusters.

The network map intelligent recommendation system with the content self-adaptive perception is further based on a clustering method of image groups: generating 3 preview images for each image layer by adopting a content-based map service preview self-adaptive sampling method, wherein the 3 preview images are expressions of information changes of different scaling proportions of the image layer, and the 3 preview images of each image layer need to be synthesized no matter clustering or image layer recommendation;

Taking 3 preview images of each layer as an image group, clustering the image groups of each layer, firstly analyzing the characteristics of the 3 preview images in the image group, wherein the 3 images come from different zoom levels of the same layer, the layers of different zoom levels express the same content of the layers under different scales, and the color expression of the same object under different scales cannot be changed;

When the image groups are clustered according to the color characteristics, the color characteristics of the image groups are synthesized, the color quantization unified histograms of the 3 images A, B, C are set as R _A,R_B,R_C respectively, the color characteristics of the image groups are R=R _A+R_B+R_C due to the similarity of the color characteristics of the images in the image groups, the characteristics of the images in the image groups are calculated and considered, and the color characteristics of the image groups are comprehensively considered.

The content self-adaptive perception network map intelligent recommendation system further comprises map service recommendation based on image groups: recommending 3 preview images of the comprehensive consideration layers based on the similar layers of the image content, and providing a similar service recommending method based on the map service sampling images on the basis;

The method comprises the steps of setting a layer D and a layer E, wherein each layer is provided with 3 preview images, namely D ₁、D₂、D₃ and E ₁、E₂、E₃ respectively, and calculating the similarity between a sampling image group of the layer D and a sampling image group of the layer E; considering that the similarity of two images is low due to the fact that the 3 images of each image layer are reduced due to the fact that similar services are positioned to different levels or different areas, the following image group similarity calculation method is proposed:

The first step: calculating the image approximations of the image groups D ₁、D₂、D₃ and E ₁、E₂、E₃ respectively, wherein the approximations are C _ji respectively by adopting a color-plus-texture feature calculation method, j= {1,2,3} is the subscript of D _j, i= {1,2,3} is the subscript of E _i, and in the 9 groups of comparison, assuming that the maximum value in C _ji is C ₁₁, D ₁ is the most similar to E ₁, and marking C ₁＝C₁₁, and excluding all groups containing D _i and E _i, namely C ₂₂、C₂₃、C₃₂、C₃₃;

And a second step of: similarly, find the maximum value in four groups C ₂₂、C₂₃、C₃₂、C₃₃, set as C ₃₂, note C ₂＝C₃₂, exclude image D ₃ and image E ₂, at this time, leave a group of images D ₂ and E ₃, note C ₃＝C₂₃;

And a third step of: after the first step and the second step, the most similar three-group image combination mode is found between the two image groups, in order to reduce the influence of different zoom levels and different positions on the similarity, different weight values k are set for each group of similarity, and the similarity between the final image groups is as follows:

C＝k₁×C₁+k₂×C₂+k₃×C₃

wherein 1 is greater than or equal to k ₁＞k₂＞k₃ is greater than or equal to 0.

The network map intelligent recommendation system with the content self-adaptive perception is further provided with a network map intelligent retrieval recommendation platform: the method comprises the steps of collecting geographic information service data, processing the data, designing and constructing a keyword search library and image feature search, constructing a Web-end network map service keyword search and similar service recommendation platform, crawling a large number of map services through a Web crawler, analyzing metadata of the map services, storing the map services into a database, constructing a keyword search library, sampling representative images for each image layer based on Apache Solr, calculating color and texture features of each image to construct a feature library, storing clustering results into a category library through clustering, and when similar images need to be recommended, finding out the category of the system, finding out the layer to which the most similar images belong under the category, and recommending the layer to a user.

Map intelligent searching recommendation platform architecture

The method comprises the steps of constructing network map services scattered on all nodes of the Internet by web crawlers, analyzing the collected structured network map service data, storing the data in a database to serve as metadata for user display, and in addition, generating thumbnails for each layer by combining an automatic generation method of network map service thumbnails through a layer acquisition interface provided by the service, wherein the thumbnails serve as service content representatives and are data bases for image feature extraction and image clustering, and the map intelligent retrieval recommendation platform framework is divided into 3 layers:

layer 1: network map service acquisition and preprocessing layer: capturing scattered map services on the Internet by compiling a web map service theme crawler, then downloading and analyzing metadata according to service versions and specifications of the map services, establishing a database table structure according to each parameter information of the map services based on a structured organization form, and storing processed data into a MySQL database;

Layer 2: image preprocessing and keyword retrieval library construction: the layer is divided into two parts, namely keyword retrieval and similar service recommendation aiming at the platform; establishing a direction index library from keywords to services aiming at keyword retrieval; generating a thumbnail for each service for similar service recommendations; the color and texture characteristics of the reserved service are calculated in advance in the background and stored in the characteristic library, so that convenience is provided for subsequent clustering;

Layer 3: platform query and recommendation logic implementation: background logic for realizing a query recommendation platform mainly comprises: and inquiring and displaying the map service according to the keywords, and recommending similar services according to the thumbnail content.

(II) network map data acquisition and processing

OGC WMSs and ARCGIS MAPSERVICE on a collection network are collected by adopting a developed geographic information service theme crawler, and ideas and logic of the geographic information service theme crawler comprise:

(1) Acquiring a seed point URL by using keywords through a search engine;

(2) Reading a seed point URL webpage by compiling a crawler program, and extracting other URL links obtained from the webpage;

(3) Analyzing the webpage content, and if the webpage contains the WMS, the web MAP SERVICE and the ARCGIS REST SERVICES keywords, reserving the webpage as a URL to be processed;

(4) If the URL link contains a service=wms & request= getcappbilities keyword and the page contains WMS-capbilities, determining that the URL link is OGC WMS service; if the link contains/arcgis/rest/services and the page content contains ARCGIS REST SERVICES directors, the service is determined to be ARCGIS SERVICE, and the service link is stored in the database.

The fetched service is stored in a database in a linked mode, metadata of the service is obtained, the request and metadata analysis are carried out on the webpage according to the service release specification type and the service version of the service, and only important information in the service release specification type and the service version of the service release specification type is selected to be stored and put in storage.

The network map intelligent recommendation system with the content self-adaptive perception comprises the following steps: establishing an inverted index library of the keywords to the services, establishing a mapping of the keywords to the service layers, and selecting two fields of description information and title of each service layer as fields matched with the keywords;

(1) The information of each layer of WMS and ARCGIS MAP SERVICE service is regarded as a Document, title and Description in the information of each layer are selected as fields of words to be segmented, the information to be indexed is handed to Tokenizer component, and the Token information of each Document is obtained through punctuation removal and pause removal operations;

(2) Then, the words are delivered to a language analysis tool, and words Term is obtained after the words obtained before are processed, including capitalization and lowercase, and words are processed into root forms;

(3) The lements Term are transmitted to an indexer Indexer, the indexed lements are established into word dictionaries and ordered according to alphabetical order, and the same lements Term are combined into a document inverted chain table Posting List;

The application aims at OGC WMS and ARCGIS MAPSERVICE map services, and finds proper service description information in metadata of the map services by carrying out keyword index on service layer information;

Establishing a search service, namely establishing a word-based inverted index table, separating English by using spaces among words and punctuation marks, and displaying information of search results, wherein the information comprises: layer header, layer description information, layer name, sample graph connection, corresponding fields are mapRiRle, mapAesc, mapMame, images respectively.

Keyword retrieval service publication: keyword retrieval is issued as a single service, the keyword retrieval service is decoupled from the whole system, tomcat is selected as an issuing server of the Solr service, after successful service issuing, information in Solr is checked through a visualized web page, a retrieval result is queried through page configuration query conditions, the service is integrated into a platform, and SolrJ tools are adopted to call the Solr search service.

Compared with the prior art, the application has the innovation points and advantages that:

firstly, clustering and image recommendation based on image content are integrated into retrieval and recommendation of network map service, so that the experience of map service search in the prior art is greatly improved, service discovery efficiency is improved, a whole set of technical processes are provided from data collection processing, thumbnail generation, image clustering and platform construction, a map service retrieval recommendation platform of a Web end is designed on the basis, the map content is accurately grasped and identified by the provided map service recommendation method based on the image content, network map intelligent recommendation can be carried out based on content adaptive perception, a map service retrieval platform based on keyword search and image content recommendation is developed, a recommendation layer has higher similarity with a comparison layer, the matching degree of the retrieval map content and actual requirements is high, the rationality and the effectiveness of the method are verified, and the method has good practical value;

Secondly, the application introduces the searching based on the image content into the network map service, develops the map service recommending function based on the image, assists the searching based on keywords, improves the searching experience of the network map service in the prior art, and provides an automatic network map service sampling method based on information quantity and approximation degree, wherein the information quantity is combined with a quadtree structure to position the region with rich content in the map, and the information quantity is approximately expressed by the complexity of the image; the approximation degree is calculated through a color quantization unified histogram and is used for screening the scaling ratio with mutation, the two scaling ratios respectively realize orderly traversal of map service from a plane position dimension and a scaling ratio dimension, effective data is provided for clustering recommendation based on image content, experiments prove that the image complexity is very effective in positioning the key position of a map, the map tile with the most abundant content under the same scaling ratio can be accurately selected, good effect is obtained by screening the key dimension by using the approximation degree, and the map service discovery efficiency is improved;

thirdly, the application provides a color feature and texture feature comprehensive representation image feature, the color feature adopts a color quantization unified histogram representation, the texture feature selects gray level co-occurrence matrix representation, and the similarity of the inclusion relation and color of an image group is considered, the image group recommendation method reduces the similarity difference caused by image layer positioning by giving different weight values to different groups of approximations, the color and texture-driven image cluster map recommendation method considers the similarity of the inclusion relation and color of the image group, and the image group recommendation method reduces the similarity difference caused by image layer positioning by giving different weight values to different groups of approximations, and experiments show that the application comprehensively considers the color and texture feature of the image group, considers the characteristics of the image group and has better recommendation results for different ground object type image layers;

Fourth, the application builds the network map intelligent retrieval recommendation platform, crawls a large number of map services through compiling a web crawler, analyzes metadata thereof and stores the map services in a database, builds a keyword retrieval library based on Apache Solr service and issues keyword retrieval service, generates and stores a thumbnail for each service, calculates the color and texture characteristics of each thumbnail and stores the thumbnail in a feature library, stores a clustering result in a category library, and finds the category of the thumbnail and then finds out the most similar image to recommend to a user under the category when recommending similar images, thereby being an important and effective supplement to the text-based service search in the prior art, making up a single geographic information service discovery mode based on the keyword search in the prior art, and having important effects and huge application value in enhancing user experience and improving the map service discovery efficiency.

Drawings

Fig. 1 is a schematic diagram of data amounts of map symbol output PNG images of different complexity.

Fig. 2 is an example diagram of locating a map complex area based on PNG data volume.

Fig. 3 is a conceptual partitioning diagram of multiple levels of a multi-source vector road network real-time fusion model.

FIG. 4 is a representation of similar color features in an image set of the present application.

Fig. 5 is a schematic diagram of a first step of screening of a map service recommendation based on an image group.

FIG. 6 is a schematic diagram of a second step of screening of image group based map service recommendations.

Fig. 7 is a schematic diagram of an input image set for experiment one of the present application.

Fig. 8 is a schematic diagram of a recommended image group for experiment one of the present application.

Fig. 9 is a schematic diagram of an input image set for experiment two of the present application.

FIG. 10 is a schematic diagram of a recommended image set for experiment two of the present application.

Fig. 11 is a schematic diagram of an input image set for experiment three of the present application.

FIG. 12 is a schematic view of a recommended image set for experiment three of the present application.

FIG. 13 is a diagram of a network map intelligent recommendation platform framework for content adaptive awareness.

FIG. 14 is a schematic diagram showing specific steps of the key index of the present application.

Fig. 15 is a schematic diagram of the map image preprocessing and image retrieval process steps.

Detailed description of the preferred embodiments

The technical scheme of the network map intelligent recommendation system with the content self-adaptive perception, which is provided by the application, is further described below with reference to the accompanying drawings, so that the application can be better understood and implemented by those skilled in the art.

The user experience of map service search and the efficiency of service discovery are of paramount importance, and image content-based searches have better visual experience and better image search efficiency than keyword text-based searches. The search function based on the image content is introduced by hundred degrees, google and the like, if the idea is introduced into the network map service, the map service recommendation function based on the image is developed, the search based on keywords can be effectively assisted, the network map service search experience in the prior art is improved, and the map service discovery efficiency is improved.

The application integrates image content-based clustering and image recommendation into network map service retrieval and recommendation to improve the map service searching experience in the prior art and improve service discovery efficiency, proposes a whole set of technical processes of data collection processing, thumbnail generation, image clustering and platform construction, and designs a map service retrieval recommendation platform for realizing a Web end on the basis, and mainly comprises the following steps:

First, a network map service automatic sampling method based on information quantity and proximity, comprising: calculating and evaluating map tile information, and calculating the approximation degree of the map tiles with the cross scales;

The method for generating the map service thumbnail based on the information quantity and the approximation degree adopts the information quantity and the quadtree structure to locate the rich-content area in the map, and the information quantity is approximately expressed by the complexity of the image; the approximation degree is calculated through a color quantization unified histogram and is used for screening scaling factors with mutation, and the scaling factors are used for orderly traversing the map service from a plane position dimension and a scaling factor dimension respectively, so that effective data is provided for clustering recommendation based on image content.

Second, color and texture driven image cluster map recommendation, comprising: map feature collaborative representation of comprehensive color and texture features, map content image clustering, an image group-based clustering method and image group-based map service recommendation;

Comprehensively representing image features by adopting color features and texture features, wherein the color features are represented by adopting a color quantization unified histogram, the texture features are represented by adopting a gray level co-occurrence matrix, and the similarity of the inclusion relation and the colors of image groups is considered, and the image group recommendation method reduces the similarity difference caused by image layer positioning by giving different weight values to the different approximation groups, so as to provide a multi-feature representation method for comprehensively representing the colors and the textures, and realize the map service recommendation based on the image groups;

Thirdly, building an intelligent retrieval recommendation platform of the network map, which comprises the following steps: the system comprises a map intelligent retrieval recommendation platform architecture, a network map data acquisition process, a map keyword retrieval library, map image preprocessing and image retrieval, and a network map intelligent retrieval recommendation platform architecture access;

A map service retrieval platform based on keyword search and image content recommendation: a large number of map services are crawled by compiling a web crawler, metadata of the map services are analyzed and stored in a database, a keyword retrieval library is built based on Apache Solr services, keyword retrieval services are issued, thumbnail images are generated for each service and stored, color and texture characteristics of each thumbnail image are calculated and stored in a characteristic library, and clustering results are stored in a category library. When recommending similar images, firstly finding the category of the similar images and then finding the most similar images under the category to recommend the most similar images to the user. The platform adopts a B/C structure, and can stably run through web pages after being released.

The application provides a map service recommending method based on image content, and develops a map service searching platform based on keyword searching and image content recommending, and the recommending layer has higher similarity with the comparing layer, thereby verifying the rationality and the effectiveness of the method.

1. Content-based map service preview self-adaptive sampling method

The map service in the prior art mostly exists in the form of map tiles, and the map tiles are small-size map images which are pre-rendered by a server side and are segmented according to a quadtree rule. For a network map service containing n levels of scaling, the minimum level of scaling L ₀ only contains 1 tile covering the whole map range, the maximum level of scaling L _n-1 contains 4n-1 map tiles, the size of the map tiles is close to the requirement of the thumbnail, compression is not needed, and flexible splicing can be performed, so that part of network map tiles are used as preview pictures, in the tile map mode, the map service concise expression is defined as that a plurality of tiles which can represent map service contents are screened out from a large number of map tiles, and the map service concise expression is further decomposed into two sub-modules: tile selection in the planar position dimension and tile selection in the scaling dimension.

(1) Tile selection for the planar position dimension: quantifying the representativeness of map tiles based on a specific scaling, for a static map, people tend to pay more attention to more complex parts such as areas with dense symbol distribution or compact mark placement, because such areas can provide more abundant information, and from this thought, the application adopts the complexity factor of the map tiles to approximately represent the information and serve as a quantification index of the representativeness of the tiles;

(2) Tile selection for the scaling dimension: the key is to find the tiles with mutation, and the map tiles with different scales have common points and differences; the common points are more, and the change among the scales is slight; the number of different points is large, and the change between the representation scales is obvious; subtle and significant correspond to subtle and abrupt changes in the map scale space, whereas abrupt changes are more noticeable to the user and should also be more incorporated into the map service thumbnail. The application adopts the image approximation degree to quantitatively evaluate the similarity degree between the two scale tiles, and judges the scaling ratio which is obviously changed based on the similarity degree.

A plurality of variables are flexibly adjusted: the number of tiles of adjacent scaling does not necessarily meet the 4-time relation, the number of tiles which affect the splicing is increased, the first r (r is more than or equal to 1) tiles with the most abundant information are used as seed tiles, and the probability of obtaining representative tiles is increased.

Map tile information amount calculation evaluation

From coarse to detailed, the number of map tiles per level of scaling grows exponentially, which is computationally intensive if all tiles are compared. According to the application, the 1 st round of screening is carried out based on the information quantity of the map, the map tiles with the most abundant information quantity are selected as seed tiles under each scaling, the geographical space range covered by the seed tiles is used as constraint heuristic type to traverse the next scaling, the maximum information quantity of the seed tiles of other levels except the L ₀ and the L ₁ levels cannot be ensured, but the higher hit rate can be obtained with lower calculation cost, the number of map tiles which participate in comparison of each level after rough selection is not more than 4, and the calculation quantity is greatly reduced;

In terms of map information amount calculation evaluation, the prior art is mostly based on statistical analysis of vector targets and is not suitable for information amount calculation of map tiles (raster images). Compared with the natural scene image, the map contains regular characters and map symbols, the number of colors is small, the regional boundary is sharp, the homogeneity of color blocks is high, and a high compression ratio can be obtained by adopting a lossless compression method. The application adopts the data quantity of the compressed map tiles as the characterization index of the information quantity. The map tiles are in PNG format, are very suitable for processing homogeneous areas in images, have higher compression ratio and can better meet the compression requirement of maps. The data volume of the PNG image is output by the map symbols with different complexity, and the data volumes of the square symbols, the round dot symbols, the annular symbols and the color symbols are sequentially increased, so that the PNG is very sensitive to image changes generated by visual variables such as shapes, textures, colors and the like, and the complexity of the map content can be represented. Fig. 2 shows an example of locating a complex area of a map based on PNG data size, each image is formed by splicing four tiles, the broken line box indicates the tile with the largest data size and requests the next scaling in this range, and it can be obtained from the three-time zooming process that the data size of the PNG image can better reflect the complexity factor of the map content. The application therefore uses the data volume of map tiles in PNG format as an information volume indicator.

(II) calculating the approximation degree of the map tiles of the cross-scale

The filtering is performed internally to each scale depending on the amount of information, and does not take into account the relationship between the different scales. After screening according to the information quantity, each scaling ratio obtains a map tile, and a map tile set is formed. The collection may contain several very similar tiles, with expression redundancy when used as a preview. The application compares the approximation degree of the map tiles between two adjacent scaling scales, judges the abrupt change or gradual change according to the approximation degree, only keeps the map tiles with obvious change, and although the maps with different scaling scales have difference in detail degree and symbol style, the overall ground object layout and distribution mode are similar, the calculation difficulty of the approximation degree of the tiles is reduced to a certain extent, the color features play a decisive role in the human map perception and cognition process as important visual variables of map symbols, so the similarity degree of the map tiles is analyzed from the color features.

Improvement 1, for color quantization: in order to consider readability and list, the types of symbols contained in the map are limited, the corresponding color quantity is also extremely lower than that of a natural scene image, however, due to the application of technologies such as anti-aliasing and the like, the edge pixels of the map symbols introduce color gradual change, the total color quantity is indirectly improved, as shown in fig. 3, map tiles in A contain 8238 colors in total, therefore, firstly, the map tiles are subjected to color quantization, and fig. 3, B, C and D are quantization results of the A with different color quantities, the visible tiles can be accurately expressed by only 16 colors, if two tiles are respectively quantized, the obtained color sets may not be consistent, the similarity of the colors per se needs to be additionally considered in the subsequent calculation of the histogram distance, and the calculation complexity is high. Firstly, splicing two tiles to be compared together, uniformly managing all colors of the two tiles, and quantifying to obtain a remarkable and unique color set;

Improvement 2, calculating the frequency of occurrence of color: the conventional method is to count the number of pixels belonging to a certain color, and the method cannot reasonably reflect the map content composition, because map symbols, especially linear symbols and punctuation symbols, usually express important information in a small number of pixels, and a histogram based on the number of pixels can greatly weaken the positions of the two types of symbols. Aiming at the problem, the map is firstly divided into a plurality of continuous color blocks, the circumferences of the color blocks are used for calculating the occurrence frequency of corresponding colors, the division standard of the continuous color blocks is that pixels with the same color are adjacent in a four-neighborhood, and a molar neighborhood boundary tracking method is adopted in circumference calculation.

In the RGB color space, the color quantization adopts a minimum variance method, and the reserved color quantity is specified. Based on the quantized color set, the occurrence frequency of each color in two tiles is counted respectively to obtain a color quantization unified histogram, and finally the similarity between the histograms is calculated by adopting cosine distance. According to the histogram of the two tiles when the color quantity is 16, the approximation degree is 0.172. The color quantity influences the value of the approximation degree, and after the experiment is tried many times, the approximation degree gradually becomes stable when the color quantity is more than the colors actually appearing in the map. Thus, subsequent experiments set a relatively high color quantity. In addition, the thematic map often contains large-area blank areas or ground color filling, and the areas with excessive weight values weaken the positions of other map symbols, so that the approximation index is interfered. The application adjusts during approximation calculation, sets the color weight value with the largest proportion in the histogram as 0, and suppresses the influence on approximation.

(III) map service preview adaptive sampling experiment analysis

The experiment selects three map services to verify the method, including a public map service and two thematic map services (medical facility distribution map and river level map). Firstly, three map tiles of the service under the minimum scaling (namely L ₀ level) are obtained as initial seed tiles, and from the representative map tiles of the automatically obtained 10 level scaling, the public map service is finally positioned in a building dense area with rich content, and the content change of the map service is better reflected in the whole amplifying process. The two thematic map services have less map tile content at maximum scale, but from the previous stages of scale, the positioning of the complex area is accurate.

The 32 colors are used for color quantization, then approximation calculation of tiles among scales is performed, and in general, two map services show different rules in the amplifying process: the mutation of the public map service occurs in the later stage, and the content of the map is obviously changed due to the introduction of elements such as blocks, buildings, greenbelts and the like; mutations in the thematic map service occur earlier because of mitigation of symbol collisions, selection of element targets, conversion of expression models, replacement of symbol patterns, and the like. Repeated research and judgment are carried out through multiple experiments, the application sets 0.1 as a critical value for thumbnail screening, and the similarity is larger than 0.1, namely, mutation is considered to occur and is brought into a thumbnail set; otherwise, not as a thumbnail.

2. Color and texture driven image cluster map recommendation

The map image is classified according to image characteristics, the image layers are clustered to reduce the searching range of similar image layers before recommending the image layers, the searching efficiency is improved, the color quantization unified histogram and the gray level co-occurrence matrix are respectively used for representing the color and texture characteristics of the image, and the image group clustering recommending method aiming at the network map service characteristics is provided on the basis of generating 3 sampling images for each image layer.

Map feature collaborative representation of integrated color and texture features

For a color map, the color features and texture features of the color map are respectively marked as G _s and G _r, for two given images D and E, the color feature difference between the two images is marked as A _s, the texture feature difference is marked as A _r,k₁ and k ₂, and the weight values of the distances between the two features are respectively marked, so that the comprehensive distance between the D and E is:

A(D，E)＝k₁*A_s(G_SD,G_SE)+k₂*A_r(G_rD,G_rE)

(II) clustering of map content images

All objects in the map data set are used as cluster centers at the beginning of clustering, each cluster center is updated through a message transmission mode in each iteration, one cluster center is selected as a center for the object which fails in competition after each iteration, a plurality of high-quality clustering results are finally obtained through continuous iteration and selection, the influence of the initial random cluster center on the final result in the partitioning method is avoided, and the result is very stable.

d_r+1(j,w)←(1-k)d_r+1(j,w)+kd_r(j，w)

the specific steps of the map content image clustering are as follows:

step 1: initializing parameters and reading map content data;

(III) clustering method based on image group

According to the method, 2355 WMS layers are selected, a content-based map service preview self-adaptive sampling method is adopted, 3 preview images are generated for each layer, the 3 preview images are representations of information changes of different scaling proportions of the layer, and the 3 preview images of each layer are required to be synthesized no matter clustering or layer recommendation.

The 3 preview images of each layer are used as an image group, the image group of each layer is clustered, the characteristics of the 3 preview images in the image group are analyzed firstly, the 3 images come from different zoom levels of the same layer, the layers with different zoom levels express the same content of the layer under different scales, the color expression of the same object under different scales cannot be changed, and most of the 3 images in the image group have similar color characteristics, as shown in fig. 4.

(IV) image group-based map service recommendation

Based on the 3 preview images of the image layers of the similar image layer recommendation comprehensive consideration image content, the application provides a similar service recommendation method based on a map service sampling image.

There are 3 preview images for each of the layers D and E, D ₁、D₂、D₃ and E ₁、E₂、E₃, respectively, to calculate the similarity between the sample image sets of the layers D and E.

Considering that the 3 images of each layer come from different zoom levels, even though layer D is similar to layer E (layers of each zoom level are very similar), during the image sampling screening process, there is a high probability that a single level of local information differences will be present for the 3 images of each layer to be sampled in different regions or at different zoom levels, which will result in a low similarity between the set of images.

Considering that the similarity of two images is low due to the fact that the 3 images of each image layer are reduced due to the fact that similar services are positioned to different levels or different areas, the following image group similarity calculation method is proposed:

The first step: calculating the image approximations of the image groups D ₁、D₂、D₃ and E ₁、E₂、E₃ respectively, wherein the approximations are C _ji respectively by adopting a color-plus-texture feature calculation method, j= {1,2,3} is the subscript of D _j, i= {1,2,3} is the subscript of E _i, and assuming that the maximum value in C _ji is C ₁₁, in the 9-group comparison, D ₁ is most similar to E ₁ and marked with C ₁＝C₁₁, and at this time, all groups containing D _i and E _i are excluded, namely C ₂₂、C₂₃、C₃₂、C₃₃ is left;

And a second step of: similarly, the maximum value found in the four groups C ₂₂、C₂₃、C₃₂、C₃₃ is set as C ₃₂ and denoted as C ₂＝C₃₂, as shown in fig. 6, excluding the images D ₃ and E ₂, and leaving the group of images D ₂ and E ₃ and denoted as C ₃＝C₂₃;

C＝k₁×C₁+k₂×C₂+k₃×C₃

(V) image clustering map recommendation experiment analysis

Based on the clustering result of the map content images, the similarity recommendation experiment is carried out by adopting the image group similarity measurement method provided by the application, and the experimental result is as follows:

experiment one: the recommendation result of the image group fig. 7 is shown in fig. 8, and the input data is map base map data and the recommendation results are map layers according to experiments, so that the image group similarity calculation method provided by the application has a good recommendation result for map planar layers. The recommended result gives consideration to both the color features and the texture features of the image group.

Experiment II: the recommended result of fig. 9 is shown in fig. 10, the input image of experiment two is a complex image, the image has complex texture and complex color, the recommended result is an image with similar texture, and the colors are similar and are all arranged in front. The image group similarity measurement method provided by the application has a good effect on complex images.

Experiment III: the recommended results of the image group in fig. 11 are shown in fig. 12, the three inputs of the experiment are layers of simple linear symbols, the recommended results are linear layers, the linear image groups with similar colors have higher similarity, and the linear image groups with large color differences are arranged in the follow-up. The image group similarity measurement method provided by the application is better recommended for the image group with linear textures.

The three groups of experiments show that the image group recommendation method aiming at the network map service characteristics comprehensively considers the color and texture characteristics of the image group and has good recommendation results for different ground object types.

3. Intelligent retrieval recommendation platform for constructing network map

The method comprises the steps of collecting geographic information service data, processing the data, designing and constructing a keyword search library and image feature search, constructing a Web-end network map service keyword search and similar service recommendation platform, crawling a large number of map services through a Web crawler, analyzing metadata of the map services, storing the map services into a database, constructing a keyword search library, sampling representative images for each image layer based on Apache Solr, calculating color and texture features of each image to construct a feature library, storing clustering results into a category library through clustering, and when similar images need to be recommended, finding out the category of the system, finding out the layer to which the most similar images belong under the category, and recommending the layer to a user.

Map intelligent searching recommendation platform architecture

The network map service scattered on each node of the Internet is collected by constructing a web crawler, the collected structured network map service data is analyzed and stored in a database to serve as metadata for user display, in addition, the thumbnail is generated for each layer by combining an automatic generation method of the network map service thumbnail through a layer acquisition interface provided by the service, the thumbnail serves as a service content representative, and the thumbnail is a data base for subsequent image feature extraction and image clustering, and the quick retrieval of the network map service and the acquisition of related map service are facilitated by constructing a retrieval recommendation platform, so that the information acquisition capability is improved. Based on this, the application proposes a map intelligent retrieval recommendation platform framework as shown in fig. 13.

The map intelligent retrieval recommendation platform framework is divided into 3 layers:

Layer 2: image preprocessing and keyword retrieval library construction: the layer is divided into two parts, namely keyword retrieval and similar service recommendation aiming at the platform; establishing a direction index library from keywords to services aiming at keyword retrieval; generating a thumbnail for each service for similar service recommendations; the application has reserved the color and texture characteristics of the service, and because the characteristic extraction and calculation are complex, the characteristics are calculated in advance in the background and stored in the characteristic library, thereby providing convenience for the subsequent clustering;

(II) network map data acquisition and processing

OGC WMSs and ARCGIS MAPSERVICE on the network are collected by adopting a developed geographic information service theme crawler, and ideas and logics of the geographic information service theme crawler are as follows:

(1) Acquiring a seed point URL by using keywords through a search engine;

The fetched service is stored in a database in a linked form, and requests and metadata analysis are carried out on the web page according to the service release specification type and the service version of the service in order to obtain the metadata of the service.

(III) map keyword search library

If the query of the key words to the service is simply matched from the database by using SQL sentences, the speed is very low, and the jam is easy to occur under the condition of high concurrency, so that the efficiency is low. Therefore, to solve these problems, an inverted index library of keywords to services is created, a mapping of keywords to service layers is created, and two fields of description information and title of each service layer are selected as fields for keyword matching. The specific steps for creating the key index are shown in fig. 14.

The application aims at two map services, namely OGC WMS and ARCGIS MAPSERVICE, and in order to index keywords of service layer information, proper service description information is found in metadata of the service layer information.

Keyword retrieval service publication: in order to relieve the high concurrency server pressure and improve the system stability, the keyword search is issued as a single service, so that the keyword search service is decoupled from the whole system, the system stability can be improved, the later system maintenance is facilitated, tomcat is selected as the issuing server of the Solr service, the information in the Solr is checked through a visual web page after the service is successfully issued, the search result is queried through page configuration query conditions, the service is integrated into a platform, a SolrJ tool is adopted, and the Solr search service is called.

(IV) map image preprocessing and image retrieval

A sampling image is generated for the OGC WMS, a respective image acquisition interface is referred, then a content-based map service previewing self-adaptive sampling method is adopted to generate a thumbnail for each service image layer, an important index for evaluating the performance of a recommendation system is the retrieval speed, the image resource data amount is large, the images in an image library are more, the sequential retrieval efficiency is low, if similar images are gathered into one type, the type of the image group of the image layer is firstly searched when the similar image layer is recommended, and then the retrieval is carried out only in the type, so that the image layer searching range can be greatly reduced.

After the image feature database is extracted, clustering operation is carried out on the color and texture features of the images, an image classification table is established, the images in each class are similar image sets, in the process of carrying out keyword query, the thumbnail of the image layer selected by a user is automatically used as an input image, the class to which the input image belongs is directly searched in a clustering result library, and then searching is carried out in the class, so that the image which is most similar to the input image or meets the specified condition is found and recommended to the user.

The map image preprocessing and image retrieval process includes the flow of fig. 15.

(V) building access of network map intelligent retrieval recommendation platform

The platform is based on a B/S architecture, adopts a Java technical stack as an implementation technology of the B/S architecture, integrates network map service management, keyword-based service retrieval and image content-based service recommendation, provides a technical scheme for integrating all services, builds a display page by using a Bootstrap technology at the front end of the platform, develops a background service by using a SPRINGMVC +spring+MyBatis integration frame of Java at the background, publishes retrieval services by using Solr and Tomcat servers as a background server, provides a keyword query interface for the background to call, stores map service metadata information by using a MySQL database, and caches service metadata information by using Redis as a cache database.

(1) Database and Redis cache

Each map service contains a plurality of map layer information, and the number of map layer information varies from a few map layers to thousands map layer information. These layer information are fixed. If the service information is displayed every time, the layer under the same service is retrieved from MySQL for display, the efficiency is necessarily low. Generating its layer information cache for each service is a solution of the present application.

The specific buffer generation time and method are as follows: when the service is accessed for the first time, searching description information of the service and layer information contained in the description information from MySQL and storing the information into Redis in json format; when the second access is performed, the service information corresponding to the service is directly obtained from Redis, so that the page response speed is improved.

(2) Map service publishing and accessing

The website background server adopts Tomcat, and the main functions are demonstrated as follows: and (5) a website entry page is searched by inputting keywords.

Claims

1. Content-adaptive web map intelligent recommendation system, characterized by integrating image content-based clustering and image recommendation into the retrieval and recommendation of web map services, from data collection and processing, thumbnail generation, image clustering, platform construction, a complete set of technical processes, and then designing and implementing a web-based map service retrieval and recommendation platform on this basis, including:

First, an automatic sampling method for web map services based on information volume and similarity includes: first, calculating and evaluating the information volume of map tiles, and second, calculating the similarity of cross-scale map tiles; using information volume combined with a quadtree structure to locate content-rich areas in the map, and the information volume is approximately expressed by image complexity; the similarity is calculated by a color quantized unified histogram to screen the scale ratio that has undergone mutations. The two methods realize the orderly traversal of map services from the plane position dimension and the scale ratio dimension respectively, providing effective data for clustering recommendations based on image content;

Second, color and texture driven image clustering map recommendation, including: first, map feature collaborative representation of integrated color and texture features, second, map content image clustering, third, clustering method based on image group, fourth, map service recommendation based on image group; color features and texture features are used to comprehensively represent image features. Color features are represented by color quantization unified histogram, and texture features are represented by grayscale co-occurrence matrix. The inclusion relationship and color similarity of image groups are considered. The image group recommendation method reduces the proximity difference caused by layer positioning by assigning different weight values to different proximity groups. A multi-feature representation method of integrated color and texture is proposed to realize map service recommendation based on image groups.

Third, build an intelligent retrieval and recommendation platform for online maps, including: 1. The map intelligent retrieval and recommendation platform architecture, 2. The network map data collection and processing, 3. The map keyword retrieval library, 4. The map image preprocessing and image retrieval, 5. The network map intelligent retrieval and recommendation platform construction and access; crawl a large number of map services by compiling web crawlers, parse their metadata and store them in the database, establish a keyword retrieval library based on the Apache Solr service and publish the keyword retrieval service, generate thumbnails for each service and store them, calculate the color and texture features of each thumbnail and store them in the feature library, store the clustering results in the category library, and when recommending similar images, first find the category to which it belongs and then find the most similar image under this category and recommend it to the user.

2. According to the content-adaptive-aware network map intelligent recommendation system of claim 1, it is characterized by a content-based map service preview adaptive sampling method: using part of the network map tiles as preview images, in the tile map mode, the map service is concisely expressed as screening out a number of tiles that can represent the map service content from a large number of map tiles, and further decomposed into two sub-modules: tile selection in the plane position dimension and tile selection in the scaling dimension;

(1) Tile selection in the plane position dimension: Based on a specific scaling ratio, the representativeness of the map tile is quantified. The complexity factor of the map tile is used to approximate its information content and is used as a quantitative indicator of tile representativeness.

(2) Tile selection in the scaling dimension: The key is to find the tile where mutation occurs. Image approximation is used to quantitatively evaluate the similarity between two scale tiles, and based on this, the scaling ratio where significant changes occur is determined.

Based on the two quantitative indicators of complexity and approximation, two rounds of screening are used to achieve full coverage of the plane position dimension and automatic selection of multi-scale tiles in the scaling dimension:

In the first round of screening, starting from the smallest scale, only the most complex tiles are selected at each level, and the geographical range of the tile is used as a constraint to process the next scale. The first round of screening filters out most tiles;

Second round of screening: Compare the similarity of tiles between scaling ratios, further remove redundancy, and define the sampling set as R. The specific steps are as follows:

Step 1: Starting from level L ₀ , add the only map tile in level L ₀ to R and use it as the seed tile;

Step 2: Request four higher-level map tiles with the same geographical range as the seed tile, and use the tile Ringo with the richest information as the new seed tile;

Step 3: Combine the four tiles to get Rpvda. If the similarity between Rpvda and the seed tile is greater than the critical value, add Rpvda to R, otherwise do not add it.

Repeat steps 2 and 3 until all scaling ratios are traversed to obtain the final set of sampling images;

Flexible adjustment of multiple variables: The number of tiles with adjacent scaling ratios does not necessarily satisfy the 4-fold relationship, which affects the number of tiles to be stitched together. The first r (r≥1) tiles with the most information are used as seed tiles to increase the probability of obtaining representative tiles.

3. The content-adaptive-aware network map intelligent recommendation system according to claim 2 is characterized in that the map tile information volume calculation and evaluation: the first round of screening is performed based on the map information volume, and the map tile with the richest information volume is selected as the seed tile at each zoom ratio, and the geographic space range covered by the seed tile is used as a constraint to heuristically traverse the next zoom ratio. Except for _L0 and _L1 levels, it is not guaranteed that the seed tiles of other levels have the maximum information volume, but a higher hit rate is obtained at a lower calculation cost. After rough selection, the number of map tiles involved in the comparison at each level does not exceed 4;

In terms of the calculation and evaluation of map information volume, the data volume of the compressed map tiles is used as the representation indicator of its information volume, and the data volume of the map tiles in PNG format is used as the information volume indicator.

4. The content-adaptive-aware network map intelligent recommendation system according to claim 1 is characterized in that the similarity of cross-scale map tiles is calculated: the similarity of map tiles between two adjacent zoom ratios is compared, and sudden or slow changes are judged accordingly, and only map tiles with significant changes are retained; the similarity of map tiles is analyzed starting from color features, and a color quantization unified histogram is used to quantify the color features of the map, and on this basis, the similarity of map tiles is measured based on the cosine distance weight;

Combining the map expression features, we make two improvements to the histogram generation method:

Improvement 1, for color quantization: First, quantize the colors of the map tiles. First, combine the two tiles to be compared together, unify all the colors of the two tiles, and quantize to obtain a significant and unique color set;

Improvement 2, calculate the frequency of color occurrence: first divide the map into several continuous color blocks, and use the perimeter of these color blocks to calculate the frequency of occurrence of the corresponding color. The division standard of continuous color blocks is that pixels of the same color are adjacent in four neighborhoods. The perimeter calculation adopts Moore neighborhood boundary tracking method;

In the RGB color space, color quantization uses the minimum variance method to specify the number of retained colors. Based on the quantized color set, the frequency of occurrence of each color in the two tiles is counted to obtain a unified color quantization histogram. Finally, the cosine distance is used to calculate the similarity between the histograms. The weight value of the color with the largest proportion in the histogram is set to 0 to suppress its influence on the approximation.

5. According to the content-adaptive-aware network map intelligent recommendation system of claim 1, it is characterized by color and texture-driven image clustering map recommendation: map images are classified according to image features, layers are clustered before layer recommendation to narrow the search range of similar layers, color quantization unified histogram and gray-level co-occurrence matrix are used to represent the color and texture features of the image, and on the basis of generating 3 sampling images for each layer, image group clustering recommendation is performed based on network map service features;

Map feature collaborative representation of integrated color and texture features: For a color map, its color feature and texture feature are recorded as _Gs and _Gr , respectively. For two given images D and E, the color feature difference between them is recorded as _As , and the texture feature difference is recorded as _Ar . _k1 and _k2 are the weight values of the distance between the two features, respectively. Then the comprehensive distance between D and E is:

A (D, E) = k ₁ *A _s (G _SD , G _SE ) + k ₂ *A _r (G _rD , G _rE )

This is used as a similarity comparison method for map images and retrieval.

6. The content-adaptive-aware network map intelligent recommendation system according to claim 1 is characterized in that, in the clustering of map content images, at the beginning of clustering, all objects in the map data set are used as cluster centers, and in each iteration, each cluster center is updated through a message passing mode, and after each iteration, the object that fails in the competition selects a cluster center as its center, and through continuous iteration and selection, multiple high-quality clustering results are finally obtained;

Clustering is performed based on the approximation between M map data points. This approximation constitutes an M×M similarity matrix. All map data points are regarded as potential cluster centers Z. The larger the value c(w, w) on the diagonal of the C matrix, the greater the probability that point w will become a cluster center. Its value is the reference degree q. The transmission mechanism includes attraction factor and attribution factor. The attraction factor and attribution factor matrices are updated. To avoid result oscillation, an attenuation coefficient k is introduced. When updating information, each information is set to k times the last iterative update value plus 1-k times the information update value. The attenuation coefficient k is a real number between 0 and 1, that is, the iterative value of t(j, w) and d(j, w) for the r+1th time:

d _r+1 (j, w) ← (1-k) d _r+1 (j, w) + kd _r (j, w)

Specific steps for map content image clustering:

Step 1: Initialize parameters and read map content data;

Step 2: Calculate the proximity matrix, find the median value of the proximity matrix and assign it to the reference degree q;

Step 3: Update the attraction factor matrix and the attribution factor matrix;

Step 4: Determine whether the maximum number of iterations has been reached or the termination condition has been met. If so, go to step 2, otherwise go to step 5;

Step 5: Get the final cluster center Z and divide each data object into the corresponding cluster.

7. According to claim 1, the content-adaptive-aware network map intelligent recommendation system is characterized by: the clustering method based on the image group: adopting the content-based map service preview adaptive sampling method, generating 3 preview images for each layer, the 3 preview images are expressions of the changes in the information of different zoom ratios of this layer, whether it is clustering or layer recommendation, it is necessary to integrate the 3 preview images of each layer;

The three preview images of each layer are taken as an image group. The image groups of each layer are clustered. First, the features of the three preview images in the image group are analyzed. The three images come from different zoom levels of the same layer. Layers at different zoom levels express the same content of the layer at different scales, and the color expression of the same object at different scales will not change.

When clustering image groups according to color features, the color features of the image group are integrated. Suppose the color quantization unified histograms of the three images A, B, and C are _RA , _RB , and _RC respectively. Due to the similarity of the color features of the images in the image group, the color feature of the image group is R= _RA + _RB + _RC . The calculation takes into account the features of the images in the image group and comprehensively considers the color features of the image group.

8. According to the content-adaptive-aware network map intelligent recommendation system of claim 1, it is characterized by the map service recommendation based on the image group: similar layer recommendation based on the image content comprehensively considers the three preview images of the layer, and on this basis, proposes a similar service recommendation method based on the map service sampling map;

Suppose there are layers D and E, each layer has three preview images, namely D ₁ , D ₂ , D ₃ and E ₁ , E ₂ , E ₃ . The similarity between the sampled image groups of layer D and layer E is calculated. Considering that the three images of each layer reduce the low similarity of the two images caused by similar services being located at different levels or different areas, the following image group similarity calculation method is proposed:

Step 1: Calculate the image similarity between image groups D ₁ , D ₂ , D ₃ and image groups E ₁ , E ₂ , E ₃ respectively, using the color plus texture feature calculation method, the similarity is C _ji , where j = {1, 2, 3} is the subscript of D _j , i = {1, 2, 3} is the subscript of E _i , assuming that the maximum value in C _ji is C ₁₁ , then in these 9 groups of comparisons, D ₁ is the most similar to E ₁ , record C ₁ = C ₁₁ , exclude all groups containing D _i and E _i , that is, C ₂₂ , C ₂₃ , C ₃₂ , C ₃₃ are left;

Step 2: Similarly, find the maximum value among the four groups C ₂₂ , C ₂₃ , C ₃₂ , and C ₃₃ , set it as C ₃₂ , record C ₂ = C ₃₂ , exclude images D ₃ and E ₂ , and now only the group of images D ₂ and E ₃ is left, record C ₃ = C ₂₃ ;

Step 3: After the first and second steps, the three most similar image combinations are found between the two image groups. To reduce the impact of different zoom levels and different positions on the similarity, a different weight value k is set for each group of similarities. The final similarity between the image groups is:

C＝k ₁ ×C ₁ +k ₂ ×C ₂ +k ₃ ×C ₃

Where 1≥k ₁ >k ₂ >k ₃ ≥0.

9. According to the content-adaptive-aware web map intelligent recommendation system described in claim 1, it is characterized by building a web map intelligent retrieval recommendation platform: including geographic information service data collection, data processing, design and construction of keyword retrieval library and image feature retrieval, building a web-based web map service keyword retrieval and similar service recommendation platform, crawling a large number of map services through web crawlers, parsing their metadata, storing them in a database and establishing a keyword retrieval library, publishing keyword retrieval services based on Apache Solr, sampling representative images for each layer, and then calculating the color and texture features of each image to build a feature library, and storing the clustering results in a category library through clustering. When it is necessary to recommend similar images, the system first finds the category to which it belongs, and then finds the layer to which the most similar image belongs under this category and recommends it to the user;

(I) Map intelligent retrieval and recommendation platform architecture

By building a web crawler to collect web map services scattered on various nodes of the Internet, the collected structured web map service data is parsed and stored in the database as metadata for display to users. In addition, through the layer acquisition interface provided by the service, combined with the automatic generation method of web map service thumbnails, thumbnails are generated for each layer. These thumbnails represent the service content and are the data basis for subsequent image feature extraction and image clustering. The map intelligent retrieval recommendation platform framework is divided into three layers:

Layer 1: Web map service collection and preprocessing layer: By compiling web map service theme crawlers, we can crawl scattered map services on the Internet, and then download and parse their metadata according to their service versions and specifications. Based on the structured organization form, we can establish a database table structure according to its various parameter information, and store the processed data in the MySQL database.

Layer 2: Image preprocessing and keyword search library construction: This layer is divided into two parts, one for the platform's keyword search and the other for similar service recommendations. For keyword search, a directional index library from keywords to services is established. For similar service recommendations, thumbnails are generated for each service. The color and texture features of the booked services are calculated in advance in the background and stored in the feature library to facilitate subsequent clustering.

Layer 3: Platform query and recommendation logic implementation: Implement the backend logic of the query recommendation platform, including: map service query and display based on keywords, and similar service recommendations based on thumbnail content;

(II) Network map data collection and processing

The geographic information service theme crawler is developed to collect OGC WMS and ArcGIS MapService on the Internet. The ideas and logic of the geographic information service theme crawler include:

(1) Obtain the seed point URL using keywords through a search engine;

(2) Read the seed URL webpage by compiling a crawler program and extract other URL links obtained from the webpage;

(3) Parse the web page content. If the page contains the keywords WMS, web map service, or ARCGIS REST Services, it will be retained as a URL to be processed.

(4) If the URL link contains the keyword service=wms&request=getcapbilities and the page contains wms-capbilities, it is determined to be an OGC WMS service; if the link contains /arcgis/rest/services and the page content contains ArcGIS REST Services Directory, it is determined to be an ArcGIS Service and the service link is stored in the database;

The captured services are stored in the database in linked form, and the metadata of the services are obtained. According to the service publishing specification type and its service version, web pages are requested and metadata parsed, and only the more important information is selected and stored in the database.

10. The content-adaptive-aware network map intelligent recommendation system according to claim 9 is characterized by: (iii) map keyword retrieval library: establishing a reverse index library of keywords to services, establishing a mapping of keywords to service layers, and selecting the description information and title fields of each service layer as the keyword matching fields;

(1) Treat each layer of WMS and ArcGIS Map Service as a document, select the Title and Description in each layer as the fields to be segmented, and pass the information to be indexed to the Tokenizer component to obtain the word-unit Token information of each document by removing punctuation and pauses;

(2) The language analysis tool is then used to process the previously obtained lemmas, including converting uppercase letters to lowercase letters and converting single words to root forms, to obtain the word Term.

(3) Pass the term to the indexer, create a word dictionary with the indexed terms and sort them in alphabetical order, merge the same terms into a posting list;

The objects are OGC WMS and ArcGIS MapService map services. The purpose is to perform keyword indexing on the service layer information and find the appropriate service description information in its metadata.

The search service is established by establishing a word-based inverted index table. In English, words are separated by spaces and punctuation marks. The search results display information including: layer title, layer description information, layer name, and sampling map connection. The corresponding fields are mapRiRle, mapAesc, mapMame, and images.

Keyword search service publishing: Publish keyword search as a separate service to decouple the keyword search service from the entire system. Select Tomcat as the publishing server for the Solr service. After the service is successfully published, view the information in Solr through a visual web page. Query the search results by configuring query conditions on the page. Integrate the service into the platform using the SolrJ tool to call the Solr search service.