CN106156063B

CN106156063B - Correlation technique and device for object picture search results ranking

Info

Publication number: CN106156063B
Application number: CN201510144467.8A
Authority: CN
Inventors: 黄恒
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2015-03-30
Filing date: 2015-03-30
Publication date: 2019-10-01
Anticipated expiration: 2035-03-30
Also published as: CN106156063A; WO2016155537A1

Abstract

The present invention provides the correlation techniques and device for object picture search results ranking, wherein the method for object picture search results ranking includes: the searching request in response to receiving user, obtain and request matched object picture with described search；Extract the visual signature of acquired object picture；Based on user to the preference score of the visual signature, user is obtained to the synthesis preference score of acquired object picture；It is object picture sequence according to synthesis preference score of the user to acquired object picture.The present invention, which realizes, can make the interest that more can accurately be close to the users to the search results ranking of object picture.

Description

Correlation technique and device for object picture search results ranking

Technical field

The present invention relates to Internet technical field more particularly to a kind of related sides for object picture search results ranking Method and device.

Background technique

With the development of internet technology, for the information content on internet also growing day by day, user is passing through search engine When searching for the information needed, often due to the search key of user's input can not information or and user required for accurate description The reasons such as the matched information content of search key is numerous cause search engine to return to a large amount of relevant information as search knot Fruit, the information content which is included is more, then user selects the difficulty of the information of oneself needs in search result Degree will be bigger.

To solve the above problems, part searches engine provides the scheme based on user interest sorted search result, specifically Are as follows: the text feature based on user's search result clusters, so that it is determined that the interested information of user, when the user it is subsequent Search result is ranked up based on user's interested information when the search engine scans for.Wherein, if a search knot The text feature content of fruit is more, then needs to carry out information extraction operation, due to being difficult to quantify in information extraction, and if Text feature is not the form types of rule, and it is difficult to extract features therein；Simultaneously because for same in different search results Attribute may use different text describing modes, for example, the attribute silk and silk of product, the really same attribute, because This, needs that search result is normalized, can not be by two identical attributes if normalized is inaccurate It recognizes on the same point of interest；In addition, since the meaning that same keyword represents under different classifications is different, for example, being directed to Keyword apple, representative meaning is different under the different classifications such as fruit, film, 3C, therefore is also needed according to up and down Literary information calculates its meaning of one's words scene and disambiguates to the keyword, if contextual information can not be obtained, disambiguating result will be by shadow It rings, then will not be inconsistent comprising the real demand largely with user according to the search result that the keyword after disambiguating scans for Information, then clustered based on the information, by the real point of interest of more difficult accurate expression user.

As it can be seen that determining that the interested information of user needs to carry out information extraction, normalizing to text feature based on text feature Change and disappear the technical treatment of the complexity such as qi, not only more difficult extraction attribute information (i.e. extraction efficiency is low), and according to being based on text The accuracy that the extracted attribute information of feature carries out user interest analysis is lower, it is more difficult to the real interest of accurate expression user Point, after being ranked up based on this to search result, user is equally more difficult to find the information really needed.

Summary of the invention

One of the technical problems solved by the invention is to provide the correlation techniques and dress for object picture search results ranking It sets, the interest that more can accurately be close to the users to the search results ranking of object picture can be made.

According to an embodiment of an aspect of the present invention, it provides a kind of for assisting object picture search results ranking Method, comprising:

It obtains user behavior within the scope of setting time and records the object picture being related to；

Extract the visual signature of acquired object picture；

Determine the number of the object picture in acquired object picture with same visual signature；

Based on the number of the object picture with same visual signature, determine user to the preference point of the visual signature Number is used for object picture search results ranking.

Optionally, the visual signature includes at least one of the following:

Object picture lightness, object picture saturation degree, object picture acutance, object picture contrast, object picture are gorgeous Degree, object picture whether picture mosaic, ratio, the object picture body color depth, object picture master shared by object picture body region Body scheme of colour.

Optionally, the number based on the object picture with same visual signature determines user to the visual signature Preference score includes:

Number and corrected value based on the object picture with same visual signature determine that the basic of the visual signature obtains Point, wherein corrected value indicates there is ratio of the object picture of the visual signature in object picture library；

Basic score according to the visual signature determines user to the preference score of the visual signature.

Optionally, the behavior record includes selection.

Optionally, the behavior record includes: browsing and selection, then based on the object picture with same visual signature Number determines that user includes: to the preference score of the visual signature

The visual signature is determined based on the number and the respective weight factor of the object picture with same visual signature Basic score, it is to browse also to choose to determination that wherein weighted factor, which is according to behavior record,；

Optionally, the visual signature is by accurate vision characteristic value.

Optionally, the visual signature is characterized by accurate vision characteristic value or level visual characteristic, wherein in vision spy In the case that sign is by level visual characteristic standard, the visual signature for extracting acquired object picture includes:

Extract the accurate vision characteristic value of the visual signature of acquired object picture；

Determine level visual characteristic belonging to the accurate vision characteristic value, the vision as acquired object picture Feature.

According to an embodiment of another aspect of the invention, a kind of side for object picture search results ranking is provided Method, comprising:

In response to receiving the searching request of user, obtains and request matched object picture with described search；

Extract the visual signature of acquired object picture；

User is obtained to the synthesis preference of acquired object picture based on preference score of the user to the visual signature Score, wherein user is according to above-mentioned for assisting the search result of object picture to the preference score of the visual signature What the method for sequence determined；

It is object picture sequence according to synthesis preference score of the user to acquired object picture.

Optionally, the visual signature includes at least one of the following:

Optionally, user is obtained to the comprehensive of acquired object picture based on preference score of the user to the visual signature Closing preference score includes:

User is inclined as synthesis of the user to acquired object picture to the sum of preference score of the visual signature Good grades.

Optionally, it obtains and requests matched object picture with described search further include: obtain and matched with described search request Object picture original ordering score, then user is obtained to acquired to the preference score of the visual signature based on user The synthesis preference score of object picture includes:

By user to the sum of preference score of the visual signature plus the original sequence point of the object picture obtained Value, as user to the synthesis preference score of acquired object picture；Or

User is added into the figure obtained to the sum of preference score of the visual signature and the product of predetermined adjusted value Obtained after the original ordering score of piece object and synthesis preference score of the value as user to acquired object picture.

It optionally, is the object picture ranked package according to synthesis preference score of the user to acquired object picture It includes:

It is the object picture to the sequence of the synthesis preference score of acquired object picture from high to low according to user Sequence.

One embodiment of another aspect according to the present invention provides a kind of for assisting object picture search results ranking Device, comprising:

Acquiring unit records the object picture being related to for obtaining user behavior within the scope of setting time；

Extraction unit, for extracting the visual signature of acquired object picture；

Number decision unit, for determining the number of the object picture in acquired object picture with same visual signature Mesh；

Preference score determination unit determines user couple for the number based on the object picture with same visual signature The preference score of the visual signature is used for object picture search results ranking.

Optionally, the visual signature includes at least one of the following:

Optionally, preference score determination unit includes:

First foundation score determines subelement, for number and correction based on the object picture with same visual signature Value determines the basic score of the visual signature, and wherein corrected value indicates the object picture with the visual signature in object picture Ratio in library；

First preference score determines subelement, for determining user to the view according to the basic score of the visual signature Feel the preference score of feature.

Optionally, the behavior record includes selection.

Optionally, the behavior record includes: browsing and selection, and preference score determination unit includes:

Second basic score determines subelement, for the number based on the object picture with same visual signature and accordingly Weighted factor determines the basic score of the visual signature, wherein weighted factor be to browse also to choose to according to behavior record Determining；

Second preference score determines subelement, for determining user to the view according to the basic score of the visual signature Feel the preference score of feature.

Optionally, the visual signature is by accurate vision characteristic value.

Optionally, the visual signature is characterized by accurate vision characteristic value or level visual characteristic, wherein in vision spy In the case that sign is by level visual characteristic characterization, extraction unit includes:

Subelement is extracted, the accurate vision characteristic value of the visual signature for extracting acquired object picture；

Level divides subelement, for determining level visual characteristic belonging to the accurate vision characteristic value, as institute The visual signature of the object picture of acquisition.

One embodiment according to a further aspect of the present invention provides a kind of dress for object picture search results ranking It sets, comprising:

Object picture acquiring unit obtains and described search request for the searching request in response to receiving user The object picture matched；

Visual Feature Retrieval Process unit, for extracting the visual signature of acquired object picture；

Comprehensive preference score obtaining unit, for obtaining user to institute based on preference score of the user to the visual signature The synthesis preference score of the object picture of acquisition, wherein user is by recited above to the preference score of the visual signature What the device for assisting the search results ranking of object picture determined；

Sequencing unit, for being object picture row according to synthesis preference score of the user to acquired object picture Sequence.

Optionally, the visual signature includes at least one of the following:

Optionally, comprehensive preference score obtaining unit is used for:

Optionally, object picture acquiring unit is also used to: being obtained and is requested the original of matched object picture with described search Ordering score, then the comprehensive preference score obtaining unit is used for:

User is added into the figure obtained to the sum of preference score of the visual signature and the product of predetermined adjusted value Obtained after the original ordering score of piece object and synthesis preference score of the value as user to acquired object picture

Optionally, sequencing unit is used for:

Picture pair with same visual signature in the object picture that the embodiment of the present application is related to based on user behavior record The number of elephant, the user determined carry out the preference of the visual signature using search result of this preference to object picture Sequence, it will be able to make the interest that more can accurately be close to the users to the search results ranking of object picture.

Since object picture can more intuitively express Properties of Objects than verbal description, based on the object picture The user that visual signature determines more can accurately be close to the users interest to the preference score of visual signature, while object picture Visual signature is easier relative to the attribute for extracting object from verbal description.

Those of ordinary skill in the art will understand that although following detailed description carries out referenced in schematic embodiment, attached drawing, But the present invention is not limited to these embodiments.But the scope of the present invention is extensive, and is intended to only through appended right It is required that limiting the scope of the invention.

Detailed description of the invention

By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, of the invention other Feature, objects and advantages will become more apparent upon:

Fig. 1 is the process of the method according to an embodiment of the invention for assisting object picture search results ranking Figure.

Fig. 2 is that the number according to an embodiment of the invention based on the object picture with same visual signature determines use Flow chart of the family to the method for the preference score of the visual signature.

Fig. 3 is the method signal of determining object picture body color depth corrected value according to an embodiment of the invention Figure.

Fig. 4 is the score schematic diagram of the determining object picture body color depth according to an embodiment of the invention.

Fig. 5 is the object picture number schematic diagram of each visual signature according to an embodiment of the invention.

Fig. 6-1 is the preference score schematic diagram of object picture lightness according to an embodiment of the invention.

Fig. 6-2 is the preference score schematic diagram of object picture acutance according to an embodiment of the invention.

Fig. 7 is that the number in accordance with another embodiment of the present invention based on the object picture with same visual signature determines Flow chart of the user to the method for the preference score of the visual signature.

Fig. 8 is the flow chart of the method according to an embodiment of the invention for object picture search results ranking.

Fig. 9-1 is the original sequence schematic diagram of search result according to an embodiment of the invention.

Fig. 9-2 is the signal according to an embodiment of the invention being ranked up according to comprehensive preference score to search result Figure.

Figure 10 is the structure of the device according to an embodiment of the invention for assisting object picture search results ranking Schematic diagram.

Figure 11 is extraction unit structural schematic diagram according to an embodiment of the invention.

Figure 12 is preference score determination unit structural schematic diagram according to an embodiment of the invention.

Figure 13 is preference score determination unit structural schematic diagram in accordance with another embodiment of the present invention.

Figure 14 is the structural representation of the device according to an embodiment of the invention for object picture search results ranking Figure.

The same or similar appended drawing reference represents the same or similar component in attached drawing.

Specific embodiment

It should be mentioned that some exemplary embodiments are described as before exemplary embodiment is discussed in greater detail The processing or method described as flow chart.Although operations are described as the processing of sequence by flow chart, therein to be permitted Multioperation can be implemented concurrently, concomitantly or simultaneously.In addition, the sequence of operations can be rearranged.When it The processing can be terminated when operation completion, it is also possible to have the additional step being not included in attached drawing.The processing It can correspond to method, function, regulation, subroutine, subprogram etc..

The computer equipment includes user equipment and the network equipment.Wherein, the user equipment includes but is not limited to electricity Brain, smart phone, PDA etc.；The network equipment includes but is not limited to that single network server, multiple network servers form Server group or the cloud consisting of a large number of computers or network servers for being based on cloud computing (Cloud Computing), wherein Cloud computing is one kind of distributed computing, a super virtual computer consisting of a loosely coupled set of computers.Its In, the computer equipment can isolated operation realize the present invention, also can access network and by with other calculating in network The present invention is realized in the interactive operation of machine equipment.Wherein, network locating for the computer equipment include but is not limited to internet, Wide area network, Metropolitan Area Network (MAN), local area network, VPN network etc..

It should be noted that the user equipment, the network equipment and network etc. are only for example, other are existing or from now on may be used The computer equipment or network that can occur such as are applicable to the present invention, should also be included within the scope of protection of the present invention, and to draw It is incorporated herein with mode.

Method (some of them are illustrated by process) discussed hereafter can be by hardware, software, firmware, centre Part, microcode, hardware description language or any combination thereof are implemented.Implement when with software, firmware, middleware or microcode When, program code or code segment to implement necessary task can be stored in machine or computer-readable medium and (for example deposit Storage media) in.Necessary task can be implemented in (one or more) processor.

Specific structure and function details disclosed herein are only representative, and are for describing the present invention show The purpose of example property embodiment.But the present invention can be implemented by many alternative forms, and be not interpreted as It is limited only by the embodiments set forth herein.

Although it should be understood that may have been used term " first ", " second " etc. herein to describe each unit, But these units should not be limited by these terms.The use of these items is only for by a unit and another unit It distinguishes.For example, without departing substantially from the range of exemplary embodiment, it is single that first unit can be referred to as second Member, and similarly second unit can be referred to as first unit.Term "and/or" used herein above include one of them or Any and all combinations of more listed associated items.

It should be understood that when a unit referred to as " connects " or when " coupled " to another unit, can directly connect Another unit is connect or be coupled to, or may exist temporary location.In contrast, " directly connect when a unit is referred to as Connect " or " direct-coupling " to another unit when, then temporary location is not present.It should explain in a comparable manner and be used to retouch State the relationship between unit other words (such as " between being in ... " compared to " between being directly in ... ", " and with ... it is adjacent Closely " compared to " with ... be directly adjacent to " etc.).

Term used herein above is not intended to limit exemplary embodiment just for the sake of description specific embodiment.Unless Context clearly refers else, otherwise singular used herein above "one", " one " also attempt to include plural number.Also answer When understanding, term " includes " and/or "comprising" used herein above provide stated feature, integer, step, operation, The presence of unit and/or component, and do not preclude the presence or addition of other one or more features, integer, step, operation, unit, Component and/or combination thereof.

It should further be mentioned that the function action being previously mentioned can be attached according to being different from some replace implementations The sequence indicated in figure occurs.For example, related function action is depended on, the two width figures shown in succession actually may be used Substantially simultaneously to execute or can execute in a reverse order sometimes.

The inventors discovered that often showing when user's search pictures object to the picture with certain class visual signature The preference of object, such as some users are specially interested in the biggish object picture of object picture lightness.Therefore, the application is implemented Example extracts the visual signature of object picture, determines user to the preference score of visual signature, so as to utilize the preference score pair The search result of object picture is ranked up, and makes the interest that more can accurately be close to the users to the search results ranking of object picture.

Present invention is further described in detail with reference to the accompanying drawing.

Fig. 1 is the process of the method according to an embodiment of the invention for assisting object picture search results ranking Figure, as shown in fig. 1, this method mainly includes the following steps:

S10, the object picture that user behavior record is related within the scope of setting time is obtained；

The visual signature of object picture acquired in S11, extraction；

The number of object picture with same visual signature in object picture acquired in S12, determination；

S13, the number based on the object picture with same visual signature determine user to the preference of the visual signature Score is used for object picture search results ranking.

Above steps is described in further detail below.

Method shown in FIG. 1 for assisting object picture search results ranking is held in user's search pictures object Capable, but the preliminary process executed before user's search pictures object in server.It can be set into periodically (such as One day, one month etc.) it is that each user executes once.In such manner, it is possible to constantly track the change of user preference according to the history of update Change, makes sequence that can more reflect the continually changing preference of user.

Firstly the need of explanation, for assisting the method for object picture search results ranking most described in the present embodiment Whole purpose is the preference score for determining user to the visual signature of object picture, so that subsequent user is searched in progress object picture Suo Shi can sort picture object search results to the preference score of the visual signature of object picture according to the user of the determination.This Application embodiment be using within the scope of setting time the user behavior object picture that is related to of record as source data, to the source data into Row feature extraction is so that it is determined that preference score of the user to the visual signature of object picture out.

Setting time described in step S10 may range from tracing forward 1 month from current time or from current time It traces 1 year forward or traces 10 years or other times range etc. forward from current time, determining for the setting time range needs Guarantee that enough source datas can be got within the scope of the setting time preference point to determine user to the visual signature Number.For example, current time is on December 31st, 2014, and it can be using one the whole month of December in 2014 as setting time range, it can also As setting time range, a year and a day in 2014 can also be regard as setting time range etc. by 10 years of 2005-2014. It is understood that the more more then identified users of the source data obtained within the scope of the setting time are to the inclined of the visual signature Good grades are more accurate, therefore, are not particularly limited here to " enough source datas ".

In the embodiment of the present application, the behavior record includes browsing and/or selection.The browsing includes but unlimited In: user, which clicks, checks that the object picture or click check corresponding product of the object picture etc..The selection includes but unlimited In: user buys the corresponding product of the object picture or user downloads the object picture or user collect the object picture or The corresponding product of the object picture is set product etc. of interest by user.

From the definition above to selection and browsing as can be seen that user is higher than the level of interest of the object picture of selection Level of interest of the user to the object picture of browsing.

In addition, it is necessary to explanation, object picture described in the present embodiment is the representative figure as a search result In the exhibiting pictures that search result list page is shown, rather than clicking the search knot in the search result list page Details picture in the details page occurred after fruit.For example, being directed to a jacket, it includes more for introducing the details picture of the jacket It is a, and showing in the picture of the search result of jacket, only show that one of represent is schemed in search result list page.I The representative figure is known as the jacket search result object picture or exhibiting pictures.Certainly, the embodiment of the present application is not arranged Except the situation that exhibiting pictures are dynamic change picture, the dynamic change is i.e. for same article or information, corresponding displaying Object picture is likely to be multiple.For example, can have figure when its suspension in search result list page for dress Piece, its be through comparison picture of effect picture and different colours on the person etc., for the situation of multiple exhibiting pictures, this Embodiment can obtain all exhibiting pictures as the object picture obtained.

Wherein, a kind of mode obtaining the object picture that user behavior record is related within the scope of setting time is as follows: passing through The historical data of the user all behaviors of website where server of User ID (mark) search server record.It can basis The object picture of first kind record identification user's selection in historical data.The user of first kind record such as server record exists Conclusion of the business record, the Download History of the website.For example, recording whether identification user buys the corresponding production of object picture according to striking a bargain Product, or according to Download History identify user whether download pictures object etc..It can be recorded and be known according to the second class in historical data The object picture of other user's browsing.The user of second class record such as server record is in the website browsing search result Ipv (browsing time of item detail page), can according to the ipv identify user whether browsing pictures object and browsing Number.By the above method can determine user browsing object picture and corresponding number and user selection object picture and Corresponding number.

Extract the visual signature of acquired object picture in step s 11, wherein the visual signature include with down toward One item missing:

It is understood that visual signature, that is, characteristics of image of object picture, each visual signature of object picture can be by Corresponding visual characteristic characterization.

The method that a kind of embodiment of the application extracts above-mentioned visual signature includes:

Object picture lightness, can by calculate object picture in HSV, (Hue tone, Saturation saturation degree, Value are bright Degree) average value of V in space obtains, and circular, which can be used, to be had any technology and realizes that the embodiment of the present application is to this It is not particularly limited.

Object picture saturation degree can be obtained, specific calculating side by calculating average value of the object picture in the S of HSV space Method, which can be used, has any technology realization, and the embodiment of the present application is not particularly limited this.

Object picture acutance carries out convolutional calculation to object picture using Laplace operator template, after convolutional calculation Obtained average value is object picture acutance, and circular, which can be used, has any technology realization, the embodiment of the present application This is not particularly limited.

The gorgeous degree of object picture: it can be by calculating object picture in the space RGB (red R, green G, indigo plant B), pixel R/G/B points The weighted average of amount and average value gap obtains, and circular, which can be used, has any technology realization, and the application is real It applies example and this is not particularly limited.

Object picture contrast: the difference of lightness (V of HSV space) and object picture lightness of each pixel of calculating can be passed through The average value of the sum of value obtains, and circular, which can be used, has any technology realization, and the embodiment of the present application does not do this Concrete restriction.

Object picture whether picture mosaic: can be identified by the continuity and tomography of picture, specific recognition methods can be used There is any technology to realize, the embodiment of the present application is not particularly limited this.Wherein, if identifying, object picture is picture mosaic, It is corresponding value be the first designated value, for example, 1 or be 101, be otherwise the second designated value, for example, 0 or be 102, this first refer to Definite value and the second designated value can be set as needed.

Ratio shared by object picture body region: due to that may include multiple entities in an object picture, it is first determined Main body in object picture (i.e. the object picture emphasis show entity) out surrounds frame can surround the minimum rectangle of each entity Area as judging means, wherein the maximum rectangle of area surrounds the main body that the entity that frame is surrounded is the object picture, meter It calculates the maximum rectangle of the area and surrounds the ratio of the area of frame and the area of entire object picture as object picture body region Shared ratio.

The object picture body color depth: the gray value by calculating main body in object picture, generally in 0-255 range It is interior, if the gray value being calculated is greater than specified gray threshold, it is determined that object picture body color is light color, if being less than or equal to The specified gray threshold, it is determined that object picture body color is dark color, wherein the specified gray threshold for example can be 127。

Object picture main body scheme of colour: the color of HSV space can be normalized to following classification: red, yellow-red, yellow It is color, green-yellow, green, blue-green, blue, purple-indigo plant, purple, red-purple, white, black, light gray, dark-grey, it identifies in object picture Ratio shared by the color composition (i.e. main body is made of which class color in above-mentioned color classification) of main body and each color is specific to count Calculation method, which can be used, has any technology realization, and the embodiment of the present application is not particularly limited this.Wherein, if it is only a kind of Ratio shared by color is greater than specified threshold, then is judged as pure color, if ratio shared by color is greater than the specified threshold in 2~3, Then it is judged as simple pattern；The ratio shared by 3 kinds or more the colors is judged as complicated pattern if being greater than the specified threshold, wherein The specified threshold for example can be 0.66.

Each visual signature for each object picture that user behavior record is related to can be obtained by step S11.

The visual signature of one of extracted object picture of embodiment is this place by accurate vision characteristic value The accurate vision characteristic value stated directly calculates the visual characteristic of acquisition by the method for extraction visual signature presented hereinbefore, Certainly, being not precluded in the present embodiment will be using the decimal point of respective rule removal visual characteristic, or reservation visual characteristic The modes such as specific bit it is special as accurate vision to the visual characteristic obtained after the visual characteristic obtained is handled is calculated Value indicative.

The visual signature of another extracted object picture of embodiment can be by accurate vision characteristic value or level vision Characteristic value.The concept of the accurate vision characteristic value is the same as described in above example.Level visual signature described herein Value is that accurate vision characteristic value is carried out level division, by each accurate vision feature value division into a level, such as one Kind division methods are as follows: corresponding 20 levels of object picture lightness are respectively the 1st~20 grade；Object picture acutance is 20 corresponding Level is the 21st~40 grade, and corresponding 20 levels of object picture saturation degree are the 41st~60 grade, and the gorgeous degree of object picture corresponds to 20 levels be the 61st~80 grade, corresponding 20 levels of object picture contrast be the 81st~100 grade, whether object picture The corresponding level of picture mosaic is the 101st~102 grade, corresponding 20 levels the 103rd~122 of ratio shared by object picture body region Grade, the corresponding level of object picture main body scheme of colour are the 123rd~125 grade, the corresponding layer of the object picture body color depth Grade is the 126th~127 grade, the corresponding accurate vision characteristic value of the corresponding specified numberical range of each level.So, for being extracted Visual signature the case where being level visual characteristic, need first to extract the accurate view of the visual signature of acquired object picture Feel characteristic value, then determine level visual characteristic belonging to the accurate vision characteristic value, by identified level visual signature It is worth the visual signature as acquired object picture.

Step S12 is for determining the number of the object picture in acquired object picture with same visual signature.Its In, after executing step S11, each visual signature for each object picture that user behavior record is related to is obtained, that is, can determine institute The number of object picture in the object picture of acquisition with same visual signature.It can specifically determine special with same vision Sign user browsing object picture number, and with same visual signature user selection object picture number.It can See, determines that the number of the object picture in acquired object picture with same visual signature determines that user behavior record relates to And the corresponding object picture number of same visual signature.

Step S13 is the number based on the object picture with same visual signature determined in step S12, to determine use Preference score of the family to the visual signature.

Wherein, it is user that a kind of embodiment, which is the object picture being related to for the user behavior record obtained in step S10, The scene of the object picture of selection determines user to the vision based on the number of the object picture with same visual signature The flow chart of the method for the preference score of feature is as shown in Figure 2, including following sub-step:

Sub-step 20 determines the visual signature based on the number and corrected value of the object picture with same visual signature Basic score；

Wherein, the corrected value indicates there is ratio of the object picture of the visual signature in object picture library.To scheme For piece objective subject shade, which includes the picture pair that object picture body color is dark color in object picture library As ratio and object picture body color shared in the object picture library are the object picture of light color in the object picture library In shared ratio.All figures that the object picture library can store for server on such as specific website, can search for The database of piece object.Method schematic diagram to determine object picture body color depth corrected value as shown in Figure 3.Assuming that figure It include 4 object pictures in piece library of object, ID range is 1000~1003, the object picture that wherein ID is 1000,1001,1003 Object picture body color be dark color, the object picture body color for the object picture that ID is 1002 is light color, then should It is that 0.75 (3/4, i.e. object picture body color is dark color that object picture body color, which is dark corrected value, in object picture library Object picture number/object picture library in object picture number), object picture body color be light color corrected value be 0.25 (1/4 and object picture body color be light color object picture number/object picture library in object picture number).Using same The method of sample can determine that the corrected value of other visual signatures.It is understood that can Fixed Time Interval for the corrected value Calculate it is primary, therefore, the not necessary step of this method the step of the determination corrected value.

Specifically, the number and corrected value based on the object picture with same visual signature determine the visual signature Basic score, can be using the number of the object picture with same visual signature and the product of corrected value inverse as vision spy The basic score of sign.For the corresponding corrected value of the object picture body color depth shown in Fig. 3, the picture of user's selection The object picture body color depth situation of object is wherein schemed in the object picture of user's selection as shown in left side table in Fig. 4 Piece objective subject color is totally 3 of dark color, and object picture body color is totally 0 of light color, the then picture pair determined according to this As shown in score right side table such as in Fig. 4 of the body color depth, object picture body color is that dark basic score is 4 (3/0.75, i.e. object picture body color is dark object picture number/corrected value), object picture body color is light color Basic score be 0 (0/0.25, i.e., object picture body color be light color object picture number/corrected value).According to this side Method can determine that the basic score of other visual signatures, be as shown in Figure 5 the object picture number of each visual signature (i.e. each view Feel the number that feature is easily selected by a user) schematic diagram, it can determine that the basic score of each visual signature according to the Fig. 5.User's selection Same visual signature object picture number it is bigger (number being easily selected by a user is more), then corresponding basic score is got over Greatly.

Sub-step 21 determines user to the preference score of the visual signature according to the basic score of the visual signature.

Existing various weighing computation methods can be used after the basic score for determining each visual signature determines user to institute The preference score of visual signature is stated, for example, the progress of LR (Logistic Regression, Logic Regression Models) model can be used Training, the embodiment of the present application are not particularly limited this.

For example, the preference score of identified each object picture lightness is as in Figure 6-1, identified each object picture is sharp As in fig. 6-2, the number that same visual signature is easily selected by a user is more for the preference score of degree, then the value of corresponding preference score It is bigger.

The application another kind embodiment is that the object picture being related to for the user behavior record obtained in step S10 includes The scene of the object picture of the object picture and user's selection of user's browsing, based on the object picture with same visual signature Number determine that user is as shown in Figure 7 to the flow chart of the method for the preference score of the visual signature, including following sub-step It is rapid:

Sub-step 70 determines the view based on the number and the respective weight factor of the object picture with same visual signature Feel the basic score of feature；

Since the object picture that user behavior record is related to includes the object picture and user that user browses in the present embodiment The object picture of selection, and user is higher than the level of interest of the object picture of browsing to the level of interest of the object picture of selection, Therefore different weighted factors can be determined for the object picture of user's browsing and the object picture of user's selection, determining vision spy When the basic score of sign, the object picture of difference user's browsing and the object picture of user's selection, respectively multiplied by corresponding weighting The factor.

For example, weighted factor can be set as 2 if behavior record is selection.If behavior record is browsing, can incite somebody to action Weighted factor is set as 0.5.For example, it is deep that user, which within the scope of setting time, has browsed 10 object picture body color depths, The object picture of color has selected 5 object picture body color depths for dark object picture, then for object picture main body Shade is dark color, basic score=10 × 0.5+5 × 2=15；User has browsed 2 pictures within the scope of setting time Objective subject shade is the object picture of light color, has selected 8 object picture body color depths for the picture pair of light color As being then light color, basic score=2 × 0.5+8 × 2=17 for the object picture body color depth.Therefore, because weighting because The effect of son, although user is to the number (15 times) for the behavior record that the object picture body color depth is dark object picture Greater than the number (10 times) to the behavior record to the object picture that the object picture body color depth is light color, but score phase Instead.

By treating the object picture of user's selection and the object picture of user's browsing with a certain discrimination, to determine the base of visual signature Plinth score can more accurately express the point of interest of user.

Sub-step 71 determines user to the preference score of the visual signature according to the basic score of the visual signature.

Existing various weighing computation methods can be used after the score for determining each visual signature determines user to the view The preference score for feeling feature, is trained for example, LR (Logistic Regression, Logic Regression Models) model can be used, The embodiment of the present application is not particularly limited this.

The object picture being related to is recorded by obtaining user behavior in the present embodiment, and the vision for extracting the object picture is special Sign, user is determined based on the number of the object picture of same visual signature to the preference score of visual signature, due to picture pair Characteristic as can more intuitively express product, and the visual signature of acquired object picture is the value of quantization, therefore is based on The user that the visual signature of the object picture determines can more accurately express the emerging of user to the preference score of visual signature It is interesting, while the visual signature of object picture is easier to extract relative to the product attribute for extracting text feature.

The method for assisting object picture search results ranking of the present embodiment can be applied to various scenes, for example, For search engine in the scene of the object picture search results ranking of user.

The embodiment of the present application also provides a kind of method for object picture search results ranking, and this method flow chart is as schemed Shown in 8, specifically comprise the following steps:

S80, the searching request in response to receiving user are obtained and are requested matched object picture with described search；

The visual signature of object picture acquired in S81, extraction；

S82, user is obtained to the synthesis of acquired object picture based on preference score of the user to the visual signature Preference score；

It S83, according to user is that the object picture sorts to the synthesis preference score of acquired object picture.

Method shown in fig. 8 for object picture search results ranking is in the case where user's search pictures object It executes, it will be by means of user determining in the method for assisting object picture search results ranking of Fig. 1 to each vision The preference score of feature.It is in response to execute in real time in the searching request of user due to the process of Fig. 8, and the process of Fig. 1 is for example It is to be periodically executed in server background, the user determined may not correspond to searching the preference score of each visual signature Real-time user preference when rope sort result, but since the preference of user has certain stability, regularly update a user couple Sequence of the not too big influence of the preference score of each visual signature to search result.

Above steps is described in further detail below.

The present embodiment is described in above example for assisting a kind of application of the method for object picture search results ranking Scene, to occurring in this present embodiment with noun same in above example, meaning is identical to be will not be described in great detail.

Wherein, step S80 is the searching request in response to receiving user, obtains and requests matched picture with described search Object, the method that wherein the embodiment of the present application requests matched object picture with described search to acquisition are not particularly limited.Its In, can obtain first with the matched search result of searching request, then obtain the corresponding object picture of the search result.

Visual signature described in step S81 includes at least one of the following:

The method of the visual signature of above-mentioned visual signature is wherein extracted with described in above example, details are not described herein again.

It should be noted that the visual signature for extracting acquired object picture in the present embodiment is accurate vision characteristic value Or level visual characteristic, it is consistent with the visual signature obtained when determining preference score of the user to visual signature, also It is, however, it is determined that the visual signature obtained when user is to the preference score of visual signature is accurate vision characteristic value, then this step The accurate vision characteristic value of the visual signature of acquired object picture, level vision that is on the contrary then obtaining visual signature need to be extracted Characteristic value.

Step S82 is to obtain user to acquired object picture based on preference score of the user to the visual signature Comprehensive preference score.

Wherein, user can determine the preference score of the visual signature using method described in above example.

A kind of embodiment of the application obtains user to acquired figure based on preference score of the user to the visual signature The method of the synthesis preference score of piece object includes:

By taking the visual signature of object picture only includes two visual signatures of object picture lightness and object picture acutance as an example, If the object picture lightness of an object picture belongs to such as the 12nd grade in Fig. 6-1, user is the 12nd grade inclined to object picture lightness Good grades are 0.02；Corresponding user belongs to the 23rd grade to the object picture acutance of the object picture, and user is to object picture Acutance is that the 23rd grade of preference score is 0.01, then user is 0.02+0.01=to the synthesis preference score of the object picture 0.03。

The application another kind embodiment obtains the preference score of the visual signature based on the visual signature and user Obtain user includes: to the method for the synthesis preference score of acquired object picture

Method due to not using the embodiment of the present application, in site search object picture matched with searching request, net Site server can give these according to verbal description of object picture etc. as prior art when providing matched object picture One intrinsic sequence of object picture, the sequence also tend to be that the ordering score based on each object picture carries out, only The ordering score is the acquisitions such as the matching degree of keyword in verbal description and searching request based on object picture etc..It will The ordering score is known as original ordering score, that is, the original ordering score is not consider the user of the present embodiment to vision spy It is the ordering score of search result distribution in the case of the preference score of sign.The present embodiment is obtaining and the matched search of request As a result the original ordering score that matched object picture is requested with described search is obtained while；Later, by user to the view The sum of preference score of feature is felt plus the original ordering score of the object picture obtained, and obtain and value is as user couple The synthesis preference score of acquired object picture.The present embodiment considers user couple when sorting picture object search results simultaneously The preference score and search result of visual signature and the matching degree of searching request, then according to this synthesis preference score to acquisition Search result, which is ranked up, can will more promote user experience.

Since the preference fractional value that obtains according to the present embodiment is the numerical value less than 1, then preference of the user to visual signature After the sum of score adds initial value ordering score, the influence to initial value ordering score is smaller, to embody user to the inclined of visual signature Influence of the good grades to original ordering score, another embodiment of the application can multiply preference score the sum of of the user to visual signature With a predetermined adjusted value, obtained product is along with obtain after acquired original ordering score and value is as comprehensive preference Score.The predetermined adjusted value for example can be 100 or 10 etc..Assuming that original ordering score is 10, user is to visual signature The sum of preference score be 0.06, then original ordering score and user are to the sum of the preference score of visual signature addition result 10+0.06=10.06, if by user to the sum of preference score of visual signature multiplied by a predetermined adjusted value 100, Zhi Houzai It is 10+0.06*100=16 with original ordering score addition result, it is seen then that the synthesis preference score that the latter calculates is to original sequence The influence of score value is bigger.

Step S83 is the object picture ranked package according to synthesis preference score of the user to acquired object picture It includes:

It wherein, can be the figure to the sequence of the synthesis preference score of acquired object picture from high to low according to user Piece object order.

As shown in Fig. 9-1, to the original sequence schematic diagram of search result, Fig. 9-2 when not use this embodiment scheme For using, to the sequence schematic diagram of search result, which is according to the user couple determined in step S82 after this embodiment scheme The size of the synthesis preference score of acquired object picture determines.

It is understood that the method for object picture search results ranking of the present embodiment is not opened up in search result Show to being executed before user, that is, original sequence schematic diagram shown in method Fig. 9-1 using the present embodiment can't be shown To user.

The present embodiment the method can visual signature based on search result and user to the inclined of the visual signature Good grades, to determine user to the synthesis preference score of acquired object picture, and according to user to acquired picture pair The synthesis preference score of elephant is object picture sequence, since the determining user of the visual signature based on object picture is to vision The preference score of feature can more accurately express the point of interest of user, and the sequence of object picture based on this can be more It is convenient to user and searches the search result needed, disappear to mitigate user and select flow caused by the search result of needs repeatedly Consumption.

The embodiment of the present application also provide it is a kind of with above for assisting the method for object picture search results ranking corresponding For assisting the device of object picture search results ranking, the apparatus structure schematic diagram is as shown in Figure 10, and the device is main Include:

Acquiring unit 100 records the object picture being related to for obtaining user behavior within the scope of setting time；

Extraction unit 101, for extracting the visual signature of acquired object picture；

Number decision unit 102, for determining the object picture in acquired object picture with same visual signature Number；

Preference score determination unit 103, for the number based on the object picture with same visual signature, determines user To the preference score of the visual signature, it to be used for object picture search results ranking.

The setting time may range from tracing forward 1 month from current time or trace 1 forward from current time Year traces forward 10 years or other times range etc. from current time, and determining for the setting time range needs to guarantee at this The preference score that enough source datas can be got within the scope of setting time to determine user to the visual signature.For example, Current time is on December 31st, 2014, can be using one the whole month of December in 2014 as setting time range, can also be by 2014 A year and a day in year, can also be by 10 years of 2005-2014 as setting time range etc. as setting time range.It is understood that , the source data obtained within the scope of the setting time more it is more then determined by users to the preference score of the visual signature more Accurately, therefore, " enough source datas " is not particularly limited here.

In the embodiment of the present application, the behavior record that has includes browsing and/or selection.The browsing includes but not Be limited to: user, which clicks, checks that the object picture or click check corresponding product of the object picture etc..It is described selection include but Be not limited to: user buys the corresponding product of the object picture or user downloads the object picture or user collects the picture pair As or user by the corresponding product of the object picture set product etc. of interest.

Wherein, acquiring unit 100 obtains the mode for the object picture that user behavior record is related within the scope of setting time such as Under: pass through the history number for the user all behaviors of website where server that User ID (mark) search server records According to.The object picture of identification user's selection can be recorded according to the first kind in historical data.First kind record such as server is remembered Conclusion of the business record, Download History of the user of record in the website.For example, recording whether identification user buys object picture according to striking a bargain Corresponding product, or according to Download History identify user whether download pictures object etc..It can be according to second in historical data The object picture of class record identification user's browsing.The user of second class record such as server record is in the website browsing search As a result ipv (browsing time of item detail page), can be identified according to the ipv user whether browsing pictures object and clear The number look at.It can determine the picture pair of object picture and corresponding number that user browses and user's selection by the above method As and object number.

Wherein, the visual signature of the extracted object picture of extraction unit 101 includes at least one of the following:

Object picture lightness, object picture saturation degree, object picture acutance, object picture contrast, object picture are gorgeous Degree, object picture whether picture mosaic, ratio, the object picture body color depth, object picture master shared by object picture body region Body scheme of colour.The method of the visual signature of each visual signature is extracted with institute in embodiment of the method above for extraction unit 101 It states, details are not described herein again.

The visual signature of the extracted object picture of one of embodiment extraction unit 101 is by the accurate vision list of feature values Sign, accurate vision characteristic value described herein are that the view of acquisition is directly calculated by the method for extraction visual signature presented hereinbefore Feel characteristic value, being not precluded certainly, in the present embodiment will be using the decimal point of respective rule removal visual characteristic, or reservation view Feel that the modes such as the specific bit of characteristic value are used as essence to the visual characteristic obtained after the visual characteristic obtained is handled is calculated True visual characteristic.

Another visual signature of the extracted object picture of embodiment extraction unit 101, by accurate vision characteristic value or Level visual characteristic characterization, the concept of the accurate vision characteristic value is the same as described in above example.Level described herein Visual characteristic is that accurate vision characteristic value is carried out level division, by each accurate vision feature value division a to level In, such as a kind of division methods are as follows: corresponding 20 levels of object picture lightness are respectively the 1st~20 grade；Object picture acutance Corresponding 20 levels are the 21st~40 grade, and corresponding 20 levels of object picture saturation degree are the 41st~60 grade, object picture Corresponding 20 levels of gorgeous degree are the 61st~80 grade, and corresponding 20 levels of object picture contrast are the 81st~100 grade, figure Whether the corresponding level of picture mosaic is the 101st~102 grade to piece object, corresponding 20 levels of ratio shared by object picture body region 103rd~122 grade, the corresponding level of object picture main body scheme of colour is the 123rd~125 grade, and object picture body color is deep Shallow corresponding level is the 126th~127 grade, the corresponding accurate vision characteristic value of the corresponding specified numberical range of each level.So, For the extracted visual signature of extraction unit 101 be level visual characteristic the case where, extraction unit 101 as shown in figure 11 can Include:

Subelement 1011 is extracted, the accurate vision characteristic value of the visual signature for extracting acquired object picture；

Level divides subelement 1012, for determining level visual characteristic belonging to the accurate vision characteristic value, makees For the visual signature of acquired object picture.

Specifically, number decision unit 102 can determine the object picture of the browsing of the user with same visual signature Number, and with same visual signature user selection object picture number.As it can be seen that determining in acquired object picture The number of object picture with same visual signature is to determine that user behavior records the corresponding figure of same visual signature being related to Piece object number.

The object picture that a kind of embodiment is related to for the user behavior record that acquiring unit 100 obtains only includes user's choosing The structure of the scene for the object picture selected, preference score determination unit 103 is as shown in Figure 12, comprising:

First foundation score determines subelement 1031, for based on the object picture with same visual signature number with Corrected value determines the basic score of the visual signature, wherein the corrected value indicates the object picture with the visual signature Ratio in object picture library.By taking the object picture body color depth as an example, which includes picture in object picture library Objective subject color is that dark object picture ratio and object picture body color shared in the object picture library are shallow The object picture of color ratio shared in the object picture library.The object picture library can be to take on such as specific website The database of all object pictures that business device stores, can search for.It is as shown in Figure 3 the determining object picture body color depth The method schematic diagram of corrected value., it is assumed that it include 4 object pictures in object picture library, ID range is 1000~1003, wherein ID Object picture body color for 1000,1001,1003 object picture is dark color, the picture for the object picture that ID is 1002 Objective subject color is light color, then it is 0.75 (3/ that object picture body color, which is dark corrected value, in the object picture library 4, i.e. object picture body color is object picture number in dark object picture number/object picture library), object picture master Body color be the corrected value of light color be 0.25 (1/4 and object picture body color be light color object picture number/picture pair As object picture number in library).It can determine that the corrected value of other visual signatures using same method.It is understood that For the corrected value can Fixed Time Interval calculate it is primary.

First preference score determines subelement 1032, for determining user to institute according to the basic score of the visual signature State the preference score of visual signature.After the basic score that first foundation score determines that subelement 1031 determines each visual signature, First preference score determines that subelement 1032 can be used existing various weighing computation methods and determine user to the visual signature Preference score, be trained for example, LR (Logistic Regression, Logic Regression Models) model can be used, the application Embodiment is not particularly limited this.

Another embodiment is that the object picture that the user behavior record obtained for acquiring unit 100 is related to includes user The scene of the object picture of browsing and the object picture of user's selection, then in the structure of preference score determination unit 103 such as Figure 13 It is shown, comprising:

Second basic score determines subelement 1033, for based on the object picture with same visual signature number with The respective weight factor determines the basic score of the visual signature；It is browsing or choosing that wherein weighted factor, which is according to behavior record, It selects to determine；

Second preference score determines subelement 1034, for determining user to institute according to the basic score of the visual signature State the preference score of visual signature.

Since the object picture that user behavior record is related to includes the object picture and user that user browses in the present embodiment The object picture of selection, and user is higher than the level of interest of the object picture of browsing to the level of interest of the object picture of selection, Therefore the second basic score determines that subelement 1033 can determine for the object picture of user's browsing and the object picture of user's selection Different weighted factors, in the score for determining visual signature, the object picture of difference user's browsing and the picture of user's selection Object, respectively multiplied by corresponding weighted factor.

By treating the object picture of user's selection and the object picture of user's browsing with a certain discrimination, to determine the base of visual signature Plinth score, the point of interest for the expression user that can more prepare.

Second preference score determines that subelement 1034 can be used existing various weighing computation methods and determine user to described The preference score of visual signature is instructed for example, LR (Logistic Regression, Logic Regression Models) model can be used Practice, the embodiment of the present application is not particularly limited this.

The embodiment of the present application also provide it is a kind of with above for the corresponding use of the method for object picture search results ranking In the device of object picture search results ranking, the apparatus structure schematic diagram is as shown in Figure 14, the device mainly includes:

Object picture acquiring unit 140 is obtained and is requested with described search for the searching request in response to receiving user Matched object picture；The embodiment of the present application obtains object picture acquiring unit 140 and requests matched picture with described search The method of object is not particularly limited.Wherein it is possible to acquisition and the matched search result of searching request first, then obtain the search As a result corresponding object picture.

Visual Feature Retrieval Process unit 141, for extracting the visual signature of acquired object picture；The visual signature packet It includes at least one of following:

Object picture lightness, object picture saturation degree, object picture acutance, object picture contrast, object picture are gorgeous Degree, object picture whether picture mosaic, ratio, the object picture body color depth, object picture master shared by object picture body region Body scheme of colour.Visual Feature Retrieval Process unit 141 extracts the method for the visual signature of above-mentioned visual signature in above example Described, details are not described herein again.

It should be noted that Visual Feature Retrieval Process unit 141 extracts the vision of acquired object picture in the present embodiment It is characterized in accurate vision characteristic value or level visual characteristic, is obtained with when determining preference score of the user to visual signature Visual signature it is consistent, it is, the visual signature obtained when if it is determined that user is to the preference score of visual signature is accurate view Feel characteristic value, then this step also needs to extract the accurate vision characteristic value of the visual signature of acquired object picture, on the contrary then obtain Take the level visual characteristic of visual signature.

Comprehensive preference score obtaining unit 142, for obtaining user based on preference score of the user to the visual signature To the synthesis preference score of acquired object picture；

Wherein, user can be by above for assisting the search result of object picture to the preference score of the visual signature The device of sequence determines that details are not described herein again；

One of embodiment, comprehensive preference score obtaining unit 142 is by user to the preference score of the visual signature The sum of as user to the synthesis preference score of acquired object picture.

Another embodiment, object picture acquiring unit 140, which is also used to obtain, requests matched picture pair with described search The original ordering score of elephant.Method due to not using the embodiment of the present application, in site search and the matched picture of searching request When object, Website server can be retouched according to the text of object picture as prior art when providing matched object picture It states etc. to the one intrinsic sequence of these object pictures, which also tends to be that the ordering score based on each object picture carries out , only the ordering score is the matching degree etc. of keyword in verbal description and searching request based on object picture etc. It obtains.The ordering score is known as original ordering score, that is, the original ordering score is not consider the use of the present embodiment Family is to the ordering score in the case of the preference score of visual signature being search result distribution.The present embodiment is obtaining and the request The original ordering score that matched object picture is requested with described search is obtained while matched search result；Later, comprehensive Preference score obtaining unit 142 asks each visual signature of object picture according to preference score of the user to the visual signature The sum of weighting；Again by the weighted sum plus obtain the object picture original ordering score, obtain and value, as with Family is to the comprehensive preference score of the interest score value of acquired object picture.The present embodiment is same when sorting picture object search results When consider user to the preference score and search result of visual signature and the matching degree of searching request, then according to this interest score value Comprehensive preference score, which is ranked up the search result of acquisition, can will more promote user experience.

Since the preference fractional value that obtains according to the present embodiment is the numerical value less than 1, then preference of the user to visual signature After the sum of score adds initial value ordering score, the influence to initial value ordering score is smaller, to embody user to the inclined of visual signature Influence of the good grades to original ordering score, the comprehensive preference score obtaining unit 142 of another embodiment of the application can be by user couple For the sum of preference score of visual signature multiplied by a predetermined adjusted value, obtained product adds acquired original ordering score Obtain afterwards and value is as comprehensive preference score.The predetermined adjusted value for example can be 100 or 10 etc..Assuming that original row Sequence score value is 10, and user is 0.06 to the sum of preference score of visual signature, then original ordering score and user are to vision spy The sum of the preference score of sign addition result is 10+0.06=10.06, if preference score the sum of of the user to visual signature multiplied With a predetermined adjusted value 100, later again with original ordering score addition result be 10+0.06*100=16, it is seen then that Hou Zheji Influence of the synthesis preference score of calculation to original ordering score is bigger.

The present embodiment consider simultaneously when sorting picture object search results user to the preference score of visual signature and The matching degree of search result and searching request is then ranked up and will more can to the search result of acquisition according to this synthesis preference score Promote user experience.

Sequencing unit 143, for being the picture pair according to synthesis preference score of the user to acquired object picture As sequence.It can be wherein the figure to the sequence of the synthesis preference score of acquired object picture from high to low according to user Piece object order.

The present embodiment can visual signature based on search result and user to the preference score of the visual signature, come Determine user to the synthesis preference score of acquired object picture, and it is inclined according to synthesis of the user to acquired object picture Good grades are object picture sequence, since the determining user of the visual signature based on object picture is to the preference of visual signature Score can more accurately express the point of interest of user, and the sequence of object picture based on this can be more convenient user The search result needed is searched, so that mitigate user selects the consumption of flow caused by the search result of needs repeatedly.

It should be noted that the present invention can be carried out in the assembly of software and/or software and hardware, for example, can adopt With specific integrated circuit (ASIC), general purpose computer or any other realized similar to hardware device.In one embodiment In, software program of the invention can be executed to implement the above steps or functions by processor.Similarly, of the invention Software program (including relevant data structure) can be stored in computer readable recording medium, for example, RAM memory, Magnetic or optical driver or floppy disc and similar devices.In addition, some of the steps or functions of the present invention may be implemented in hardware, example Such as, as the circuit cooperated with processor thereby executing each step or function.

In addition, a part of the invention can be applied to computer program product, such as computer program instructions, when its quilt When computer executes, by the operation of the computer, it can call or provide according to the method for the present invention and/or technical solution. And the program instruction of method of the invention is called, it is possibly stored in fixed or moveable recording medium, and/or pass through Broadcast or the data flow in other signal-bearing mediums and transmitted, and/or be stored according to described program instruction operation In the working storage of computer equipment.Here, according to one embodiment of present invention including a device, which includes using Memory in storage computer program instructions and processor for executing program instructions, wherein when the computer program refers to When enabling by processor execution, method and/or skill of the device operation based on aforementioned multiple embodiments according to the present invention are triggered Art scheme.

It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims Variation is included in the present invention.Any reference signs in the claims should not be construed as limiting the involved claims.This Outside, it is clear that one word of " comprising " does not exclude other units or steps, and odd number is not excluded for plural number.That states in system claims is multiple Unit or device can also be implemented through software or hardware by a unit or device.The first, the second equal words are used to table Show title, and does not indicate any particular order.

Claims

1. a kind of method for assisting object picture search results ranking characterized by comprising

Extract the visual signature of acquired object picture；

Based on the number of the object picture with same visual signature, determine that user to the preference score of the visual signature, uses In object picture search results ranking；

The wherein number based on the object picture with same visual signature, determines user to the preference of the visual signature Score includes:

Number and corrected value or weighted factor based on the object picture with same visual signature determine the visual signature Basic score, wherein corrected value indicates there is ratio of the object picture of the visual signature in object picture library, and wherein It is to browse also to choose to determination that weighted factor, which is according to behavior record,；

2. the method as described in claim 1, which is characterized in that the visual signature includes at least one of the following:

Object picture lightness, object picture saturation degree, object picture acutance, object picture contrast, the gorgeous degree of object picture, figure Piece object whether picture mosaic, ratio, the object picture body color depth shared by object picture body region, the color matching of object picture main body Scheme.

3. the method as described in claim 1, which is characterized in that the behavior record includes selection.

4. the method as described in claim 1, which is characterized in that the visual signature is by accurate vision characteristic value.

5. the method as described in claim 1, which is characterized in that the visual signature is by accurate vision characteristic value or level vision Characteristic value, wherein extracting acquired object picture in the case where visual signature is by level visual characteristic standard Visual signature includes:

Determine level visual characteristic belonging to the accurate vision characteristic value, the vision as acquired object picture is special Sign.

6. a kind of method for object picture search results ranking characterized by comprising

Extract the visual signature of acquired object picture；

User is obtained to the synthesis preference score of acquired object picture to the preference score of the visual signature based on user, Wherein, user is according to described in any one of claim 1 to 5 for assisting picture pair to the preference score of the visual signature What the method for the search results ranking of elephant determined；

7. method as claimed in claim 6, which is characterized in that the visual signature includes at least one of the following:

8. method as claimed in claim 6, which is characterized in that used based on preference score of the user to the visual signature Family includes: to the synthesis preference score of acquired object picture

User divides the sum of preference score of the visual signature as synthesis preference of the user to acquired object picture Number.

9. method as claimed in claim 6, which is characterized in that obtain and request matched object picture also to wrap with described search It includes: obtaining the original ordering score for requesting matched object picture with described search, then based on user to the visual signature Preference score obtains user

User makees the sum of preference score of the visual signature plus the original ordering score of the object picture obtained It is user to the synthesis preference score of acquired object picture；Or

User is added into the picture pair obtained to the sum of preference score of the visual signature and the product of predetermined adjusted value Obtained after the original ordering score of elephant and synthesis preference score of the value as user to acquired object picture.

10. the method as described in any of claim 6,8,9, which is characterized in that according to user to acquired picture pair The synthesis preference score of elephant is that object picture sequence includes:

It is that the object picture sorts to the sequence of the synthesis preference score of acquired object picture from high to low according to user.

11. a kind of for assisting the device of object picture search results ranking characterized by comprising

Number decision unit, for determining the number of the object picture in acquired object picture with same visual signature；

Preference score determination unit determines user to described for the number based on the object picture with same visual signature The preference score of visual signature is used for object picture search results ranking；

Preference score determination unit includes:

Basic score determines subelement, for number and corrected value or weighting based on the object picture with same visual signature The factor determines the basic score of the visual signature, and wherein corrected value indicates the object picture with the visual signature in picture pair As the ratio in library, it is to browse also to choose to determination that wherein weighted factor, which is according to behavior record,；

Preference score determines subelement, for determining user to the visual signature according to the basic score of the visual signature Preference score.

12. device as claimed in claim 11, which is characterized in that the visual signature includes at least one of the following:

13. device as claimed in claim 11, which is characterized in that the behavior record includes selection.

14. device as claimed in claim 11, which is characterized in that the visual signature is by accurate vision characteristic value.

15. device as claimed in claim 11, which is characterized in that the visual signature is regarded by accurate vision characteristic value or level Characteristic value is felt, wherein extraction unit includes: in the case where visual signature is characterized by level visual characteristic

Level divides subelement, for determining level visual characteristic belonging to the accurate vision characteristic value, as acquired Object picture visual signature.

16. a kind of device for object picture search results ranking characterized by comprising

Object picture acquiring unit obtains matched with described search request for the searching request in response to receiving user Object picture；

Comprehensive preference score obtaining unit, for obtaining user to acquired based on preference score of the user to the visual signature Object picture synthesis preference score, wherein user is by claim 11 to 15 to the preference score of the visual signature What the device of described in any item search results rankings for assisting object picture determined；

Sequencing unit, for being object picture sequence according to synthesis preference score of the user to acquired object picture.

17. device as claimed in claim 16, which is characterized in that the visual signature includes at least one of the following:

18. device as claimed in claim 16, which is characterized in that comprehensive preference score obtaining unit is used for:

19. device as claimed in claim 16, which is characterized in that object picture acquiring unit is also used to: acquisition is searched with described Rope requests the original ordering score of matched object picture, then the comprehensive preference score obtaining unit is used for:

20. the device as described in any of claim 16,18,19, which is characterized in that sequencing unit is used for: