CN113537089B - Pine wood nematode epidemic wood identification and positioning method based on unmanned aerial vehicle aerial photography original sheet - Google Patents
Pine wood nematode epidemic wood identification and positioning method based on unmanned aerial vehicle aerial photography original sheet Download PDFInfo
- Publication number
- CN113537089B CN113537089B CN202110820396.4A CN202110820396A CN113537089B CN 113537089 B CN113537089 B CN 113537089B CN 202110820396 A CN202110820396 A CN 202110820396A CN 113537089 B CN113537089 B CN 113537089B
- Authority
- CN
- China
- Prior art keywords
- epidemic
- wood
- image
- identification
- special mark
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/587—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using geographical or spatial information, e.g. location
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Library & Information Science (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of pine wood nematode disease image recognition and positioning, and discloses a pine wood nematode disease epidemic wood recognition and positioning method based on an unmanned aerial vehicle aerial original sheet, wherein after epidemic wood is recognized on the original sheet, special marks which are designed by self are embedded in epidemic wood positions, the special marks have obvious appearance characteristics in mountain forests, are not easy to be confused with other objects, and can still keep the characteristics after orthographic images are spliced; the detection and identification of the special marks are carried out on the orthographic image with the accurate longitude and latitude information, so that the longitude and latitude position of each special mark can be accurately obtained, the quantity and distribution of the epidemic woods in the target mountain forest area range can be accurately counted, and the identification and positioning of the pine wood nematode epidemic woods of the plant can be accurately realized.
Description
Technical Field
The invention relates to the technical field of identification and positioning of pine wood nematode disease images, in particular to a pine wood nematode disease wood identification and positioning method based on an unmanned aerial vehicle aerial photo original sheet.
Background
Pine wood nematode disease, also called pine tree wilt disease, is a destructive forest disease caused by pine wood nematodes, pine tree after being infected by the pine wood nematode disease, needle leaves are yellow brown or reddish brown, wilt and droops, resin secretion is stopped, gradual withered and dead, and finally rotten, and the pine wood nematodes have strong pathogenicity, fast death speed of hosts, quick spread, large mountain forest area, difficult finding in time and great treatment difficulty.
Because the infected pine wood nematode epidemic wood is yellow brown or reddish brown, needle leaves are gradually withered, and the appearance is different from that of normal pine trees and other tree species in mountain forests, a solution for acquiring mountain forest images by using an unmanned aerial vehicle aerial photographing technology and intelligently identifying the epidemic wood based on a deep learning technology also exists at present.
Existing solutions are mainly of two types:
(1) And (3) directly carrying out epidemic wood identification on the unmanned aerial vehicle aerial photo original sheet. According to the scheme, an unmanned aerial vehicle is utilized to obtain an aerial image of a target mountain forest area, and various deep learning algorithm models are used for directly detecting and identifying pine wood nematode epidemic on an aerial image raw sheet.
The technical scheme for identifying epidemic wood is directly carried out on an unmanned aerial vehicle aerial photo raw sheet: the unmanned aerial vehicle is when gathering mountain forest image, the Shan Zhangyuan pieces of shooting also can obtain the longitude and latitude information of unmanned aerial vehicle position when this former piece of shooting according to GPS signal, nevertheless because receive unmanned aerial vehicle flight angle route planning and the influence of wind direction at that time, this longitude and latitude coordinate can't pinpoint to the positive central point position of this image, simultaneously because mountain forest topography is uneven, can't be accurate carry out the one-to-one with every pixel position and longitude and latitude coordinate in the former piece, be difficult to carry out accurate longitude and latitude location to the epidemic timber that is identified, with the unable accurate matching of plant epidemic timber in the adjacent former piece remove the weight, consequently although epidemic timber recognition effect is better, nevertheless can't accomplish and carry out accurate quantity and distribution statistics to the epidemic timber in the regional scope.
(2) And performing epidemic wood identification and positioning on the orthographic image attached with longitude and latitude coordinate information after splicing. According to the scheme, firstly, aerial images of mountain forests are acquired by using an unmanned aerial vehicle, larger overlapping exists between adjacent aerial image raw sheets, phase control points are arranged on the ground at intervals, after the aerial image raw sheets of the unmanned aerial vehicle are acquired, the raw sheets are spliced into a large-range mountain forest area orthoimage with longitude and latitude coordinate information by using image splicing software or an algorithm, each pixel point in the orthogram can acquire accurate longitude and latitude coordinate information, then, the deep learning algorithm model is used for identifying and positioning pine wood nematode epidemic on the spliced orthoimage, and the quantity and distribution condition of the pine wood nematode epidemic in the range of a target mountain forest are counted.
The technical scheme for identifying and positioning on the spliced orthographic image has the defects that: in the process of matching and splicing the original unmanned aerial vehicle aerial photo, due to the influences of different shooting positions of adjacent original unmanned aerial vehicle aerial photo, visual angle distortion, insufficient precision in matching, leaf shaking caused by external wind direction and the like, the phenomenon of virtual shadow, blurring and the like exists in the orthographic image generated by splicing, and the image quality is reduced to a certain extent compared with the original unmanned aerial photo, so that the scheme can accurately position the longitude and latitude of each pixel position in the image, but has a larger influence on the identification accuracy of epidemic wood.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a pine wood nematode disease wood identification and positioning method based on an unmanned aerial vehicle aerial photo original sheet, which is used for realizing the identification and positioning of the pine wood nematode disease wood of a plant in an ultra-large mountain forest area accurately by embedding a special mark in an original disease wood position after the aerial photo original sheet is identified and identifying and positioning the special mark in an orthographic image after splicing so as to solve the problems.
The technical scheme adopted for solving the technical problems is as follows: a pine wood nematode disease wood identifying and positioning method based on an unmanned aerial vehicle aerial photo original sheet comprises the following steps:
(1) Collecting aerial image raw sheets in a mountain forest area range through an aerial camera carrying mode of an unmanned aerial vehicle, selecting an image of which part contains pine wire insect epidemic wood as an epidemic wood training sample, and then labeling epidemic wood in the epidemic wood training sample image through an image labeling tool;
(2) Selecting a proper object detection model based on a deep learning technology as a epidemic wood recognition model, and training the epidemic wood recognition model by using the epidemic wood training sample generated in the step (1);
(3) Performing epidemic detection and identification on the images of all the aerial image raw sheets by using the epidemic identification model trained in the step (2) to obtain all the epidemic and the position information of all the epidemic in the images of each aerial image raw sheet;
(4) Embedding special marks with obvious features at epidemic wood positions of the images of each aerial image original piece obtained in the step (3), and splicing all aerial image original pieces embedded with the special marks into an orthographic image with longitude and latitude coordinates by using image splicing software or an algorithm;
(5) Selecting an image of which part contains a special mark in the orthographic image as a special mark training sample, and marking the special mark in the special mark training sample image by an image marking tool;
(6) Selecting a proper object detection model based on a deep learning technology as a special mark recognition model, and training the special mark recognition model by using the special mark training sample generated in the step (5);
(7) Performing special mark recognition on all the orthographic images by using the special mark recognition model trained in the step (6) to obtain all the special marks and the position information thereof in the orthographic images;
(8) And (3) converting the coordinate information of the special mark generated in the step (7) into longitude and latitude coordinates, and acquiring the number of epidemic woods in the target mountain range and the longitude and latitude of each plant of epidemic woods, thereby realizing the identification and positioning of the plant of epidemic woods accurately.
Preferably, the method for obtaining the epidemic wood training sample in the step (1) includes the steps of determining a target mountain forest area range in advance, planning and setting a route, setting a plurality of phase control points on the ground, enabling the unmanned aerial vehicle to fly at a constant speed at a fixed altitude, shooting a mountain forest area original sheet by an unmanned aerial vehicle carrying camera every fixed time to obtain all original sheets, uniformly cutting each original sheet into 768 x 768 pixel small images, randomly selecting partial original sheet cut small images, and manually selecting all epidemic wood in the images by using an image marking tool labelImg to obtain the epidemic wood training sample.
Preferably, the epidemic wood recognition model in the step (2) selects YOLOv object detection model as the standard model of epidemic wood recognition, the value of the last full-connection layer of the YOLOv object detection model is set to 2, and the model is used for representing two epidemic wood types of current year epidemic wood and past year epidemic wood, and all epidemic wood training samples are represented by 4:1:1 is divided into a training set, a verification set and a test set, the loss function is calculated through the training set, the model parameters are updated through back propagation, the super parameters of the model are optimized through the verification set, and finally the test set is used for testing the identification accuracy of the model under the epidemic training sample data set.
Preferably, the method for detecting and identifying epidemic woods in the step (3) includes that all aerial image raw sheets shot by an unmanned aerial vehicle are segmented, each raw sheet is evenly segmented into 768 x 768 pixel small images, the small images are input into a trained epidemic woods identification model to carry out epidemic woods identification, the epidemic woods identification model is output into pixel positions, epidemic woods types and confidence degrees of epidemic woods identification frames in the small images, and finally positions of all the epidemic woods identification frames belonging to the raw sheet are converted into positions of the raw sheet according to the positions of each small image in the raw sheet.
Preferably, the special mark with obvious embedded features in the step (4) is a white solid circle with a radius of 20 pixels.
Preferably, in the step (4), all original sheets embedded with special marks are led into image stitching software or algorithm for stitching, and an orthographic image with longitude and latitude coordinates is generated, wherein each pixel point in the orthographic image corresponds to one longitude and latitude coordinate.
Preferably, the special mark training sample in the step (5) is obtained by segmenting the obtained orthographic image into 768×768 pixels of small images, and randomly selecting a plurality of small images containing special marks as the special mark training sample, wherein the image marking tool is labelImg marking tools.
Preferably, the special mark recognition model YOLOv in the step (6) is selected as a special mark recognition model, and the value of the last full-connection layer of the special mark recognition model is set to be 1.
Preferably, the method for identifying the special marks of all the orthographic images in the step (7) includes inputting the segmented orthographic image small images into a trained special mark identification model for identification, outputting the special mark identification model as the pixel positions of the special marks in the small images and the confidence degrees, namely the pixel positions of the epidemic woods in the small images, setting the confidence degree threshold to be 0.8, and finally reserving the special mark identification result with the confidence degree exceeding the threshold.
Preferably, the method for converting the coordinate of longitude and latitude in the step (8) includes obtaining the pixel positions of all the center points of the special marks in the orthographic image according to the pixel positions of the small images in the orthographic image and the pixel positions of the special marks in each small image, and obtaining the longitude and latitude coordinates of the center points through the corresponding relation between the pixel positions and the longitude and latitude coordinates.
Compared with the prior art, the invention has the beneficial effects that:
1. Epidemic wood detection and identification are directly carried out on an original aerial image sheet of the unmanned aerial vehicle, and the accuracy of identifying the epidemic wood is higher due to higher image quality of the original aerial image sheet.
2. After epidemic wood is identified on the original sheet, a self-designed special mark is embedded in the epidemic wood position, the special mark has obvious appearance characteristics in mountain forests, is not easy to be confused with other objects, and can still keep the characteristics after orthographic images are spliced.
3. The detection and identification of the special marks are carried out on the orthographic image with the accurate longitude and latitude information, so that the longitude and latitude position of each special mark can be accurately obtained, the quantity and distribution of the epidemic woods in the target mountain forest area range can be accurately counted, and the identification and positioning of the pine wood nematode epidemic woods of the plant can be accurately realized.
Drawings
FIG. 1 is a flow chart of the present invention.
Fig. 2 is a schematic diagram of epidemic wood detection and identification of an image of an aerial image original sheet.
FIG. 3 is a schematic illustration of an orthographic image stitched after embedding a special mark.
Fig. 4 is a schematic diagram of the obtained longitude and latitude coordinate data information.
Detailed Description
The following describes the embodiments of the present invention further with reference to the drawings. The description of these embodiments is provided to assist understanding of the present invention, but is not intended to limit the present invention. In addition, the technical features of the embodiments of the present invention described below may be combined with each other as long as they do not collide with each other.
Referring to fig. 1, a pine wood nematode disease wood identifying and positioning method based on an unmanned aerial vehicle aerial photo original sheet comprises the following steps:
(1) Collecting aerial image raw sheets in a mountain forest area range through an aerial camera carrying mode of an unmanned aerial vehicle, selecting an image of which part contains pine wire insect epidemic wood as an epidemic wood training sample, and then labeling epidemic wood in the epidemic wood training sample image through an image labeling tool;
(2) Selecting a proper object detection model based on a deep learning technology as a epidemic wood recognition model, and training the epidemic wood recognition model by using the epidemic wood training sample generated in the step (1);
(3) Performing epidemic detection and identification on the images of all the aerial image raw sheets by using the epidemic identification model trained in the step (2) to obtain all the epidemic and the position information of all the epidemic in the images of each aerial image raw sheet;
(4) Embedding special marks with obvious features at epidemic wood positions of the images of each aerial image original piece obtained in the step (3), and splicing all aerial image original pieces embedded with the special marks into an orthographic image with longitude and latitude coordinates by using image splicing software or an algorithm;
(5) Selecting an image of which part contains a special mark in the orthographic image as a special mark training sample, and marking the special mark in the special mark training sample image by an image marking tool;
(6) Selecting a proper object detection model based on a deep learning technology as a special mark recognition model, and training the special mark recognition model by using the special mark training sample generated in the step (5);
(7) Performing special mark recognition on all the orthographic images by using the special mark recognition model trained in the step (6) to obtain all the special marks and the position information thereof in the orthographic images;
(8) And (3) converting the coordinate information of the special mark generated in the step (7) into longitude and latitude coordinates, and acquiring the number of epidemic woods in the target mountain range and the longitude and latitude of each plant of epidemic woods, thereby realizing the identification and positioning of the plant of epidemic woods accurately.
The method for acquiring the epidemic wood training sample in the step (1) comprises the steps of presetting a target mountain forest area range, planning and setting a route, setting a plurality of phase control points on the ground, enabling an unmanned aerial vehicle to fly at a constant altitude at a constant speed, shooting a mountain forest area original sheet by an unmanned aerial vehicle carrying camera every fixed time to acquire all original sheets, and enabling subsequent orthographic image splicing to have higher precision, enabling adjacent original sheets to have larger course overlapping rate, and enabling adjacent routes to have certain side overlapping rate; after all the original sheets are obtained, each original sheet is uniformly cut into a small image with 768×768 pixels, then partial small images cut from the original sheets are randomly selected, and an image marking tool is utilized.
The epidemic wood recognition model in the step (2) selects YOLOv object detection models with higher recognition accuracy and higher recognition speed as reference models for epidemic wood recognition, and in order to enable the epidemic wood training samples to have diversity and wider representativeness, the training pictures are subjected to a data augmentation strategy based on color space change and shape change, and the method mainly comprises random change in contrast and saturation, random rotation, random left-right overturn, random up-down overturn and random scale scaling; because epidemic wood training samples are classified into new epidemic wood and past epidemic wood in the current year, the value of the last full-connection layer of YOLOv object detection models is set to be 2, and the full-connection layer is used for representing the new epidemic wood and the past epidemic wood in the current year, and all the epidemic wood training samples are represented by 4:1:1 is divided into a training set, a verification set and a test set, the loss function is calculated through the training set, the model parameters are updated through back propagation, the super parameters of the model are optimized through the verification set, and finally the test set is used for testing the identification accuracy of the epidemic wood identification model under the epidemic wood training sample data set.
As shown in fig. 2, the method for detecting and identifying epidemic woods in the step (3) includes that all aerial image raw sheets shot by an unmanned aerial vehicle are segmented, each raw sheet is evenly segmented into 768 x 768 pixel small images, the small images are input into a trained epidemic woods identification model for epidemic woods identification, the epidemic woods identification model is output as pixel positions, epidemic woods types and confidence degrees of epidemic woods identification frames in the small images, and finally positions of all the epidemic woods identification frames belonging to the raw sheet are converted into positions of the raw sheet according to the positions of each small image in the raw sheet.
Referring to an orthographic image schematic diagram spliced by embedding special marks in fig. 3, in the step (4), special marks with obvious features are embedded in epidemic wood positions of images of original sheets of each aerial image, and because of more overlapped parts in adjacent original sheets, a condition that one epidemic wood is repeatedly identified for many times exists, and Shan Zhangyuan sheets cannot achieve accurate longitude and latitude positioning, so that duplicate removal is difficult to be directly carried out on the original sheets; after the epidemic wood pixel positions are identified on the original sheet, embedding special marks into the center points of all the identification frames; the choice of the special mark is taken into account: firstly, the characteristics are obvious in mountain forest areas, and obvious differences exist between the characteristics and the appearance of various existing objects; secondly, even if the special mark caused in the process of splicing the original sheet into the orthographic image is subjected to various stretching deformation and cutting-off, the normal identification can still be realized; thirdly, the embedded special mark cannot obviously influence the registration and splicing of the images. Considering the above 3 points, the special mark with obvious embedded features is selected as a white solid circle with a radius of 20 pixels, and it should be noted that any other object meeting the three requirements can be selected as the special mark. All original sheets embedded with special marks are led into image splicing software or algorithm for splicing, an orthographic image with longitude and latitude coordinates is generated, the coverage range of the orthographic image can reach hundreds of square kilometers, the pixel value is hundreds of billions of pixels, and each pixel point in the orthographic image corresponds to one longitude and latitude coordinate.
The special mark training sample in the step (5) is obtained by dividing the obtained orthographic image into 768 x 768 pixel small images, and randomly selecting a plurality of small images containing the special mark as the special mark training sample, wherein distortion correction of an original sheet can exist in the orthographic image splicing process, a certain epidemic strain is divided into different small images in the original sheet small image dividing process, the special mark at the splicing position has deformation and incompleteness to a certain extent, and various scenes can be contained as far as possible when the labeling sample is selected; the image marking tool is labelImg marking tool, and all special marks in the sample picture are manually framed.
The special mark recognition model in the step (6) is selected YOLOv model as the special mark recognition model, in the step (5), the special mark has shape change in the splicing process, and the data of the image is amplified before training, and the amplifying strategy mainly comprises random rotation, random left-right overturn, random up-down overturn and random stretching; because the special mark is only selected in one type, the value of the last full-connection layer of the special mark identification model is set to be 1; the data set dividing mode is similar to the model parameter training process and training epidemic models in the step (2), namely the training epidemic models are divided into a training set, a verification set and a test set, the loss function is calculated through the training set, the model parameters are updated through back propagation, the super parameters of the models are adjusted through the verification set, and finally the test set is used for testing the identification accuracy of the special mark identification model under the special mark training sample data set.
The method for carrying out special mark recognition on all the orthographic images in the step (7) comprises the steps of inputting the segmented orthographic image small images into a trained special mark recognition model for recognition, outputting the special mark recognition model into pixel positions and confidence degrees of special marks in the small images, namely pixel positions of epidemic woods in all the small images, setting a confidence coefficient threshold value to be 0.8, finally reserving a special mark recognition result with the confidence coefficient exceeding the threshold value, and reducing the false recognition rate of the special mark recognition model.
The method for converting into longitude and latitude coordinates in the step (8) comprises the steps of obtaining pixel positions of all special mark center points in an orthographic image according to pixel positions of small images in the orthographic image and pixel positions of special marks in each small image, and obtaining longitude and latitude coordinates of the center points through corresponding relations between the pixel positions and the longitude and latitude coordinates, so that epidemic wood quantity in a target mountain range and longitude and latitude coordinate distribution of each plant of epidemic wood are obtained, and accurate identification and positioning of the plant of epidemic wood are achieved, wherein each row of data in the image comprises 4 pieces of information, each piece of information is separated by space symbols, and epidemic wood types (1 represents new epidemic wood in the current year, 2 represents epidemic wood in the current year), longitudes, latitudes and confidence degrees are respectively obtained from left to right, and the higher the confidence degrees represent that the model considers that the place is epidemic wood more likely.
According to the invention, epidemic wood detection and identification are directly carried out on the original film of the aerial image of the unmanned aerial vehicle, and the accuracy of identifying the epidemic wood is higher due to higher image quality of the original film; after epidemic wood is identified on the original sheet, embedding a self-designed special mark at the epidemic wood position, wherein the special mark has obvious appearance characteristics in mountain forests, is not easy to be confused with other objects, and can still keep the characteristics after orthographic images are spliced; the detection and identification of the special marks are carried out on the orthographic image with the accurate longitude and latitude information, so that the longitude and latitude position of each special mark can be accurately obtained, the quantity and distribution of the epidemic woods in the target mountain forest area range can be accurately counted, and the identification and positioning of the pine wood nematode epidemic woods of the plant can be accurately realized.
The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. It will be apparent to those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, and yet fall within the scope of the invention.
Claims (9)
1. A pine wood nematode disease wood identifying and positioning method based on an unmanned aerial vehicle aerial photo original sheet is characterized by comprising the following steps:
(1) Collecting aerial image raw sheets in a mountain forest area range through an aerial camera carrying mode of an unmanned aerial vehicle, selecting an image of which part contains pine wire insect epidemic wood as an epidemic wood training sample, and then labeling epidemic wood in the epidemic wood training sample image through an image labeling tool;
(2) Selecting an object detection model based on a deep learning technology as a epidemic wood recognition model, and training the epidemic wood recognition model by using the epidemic wood training sample generated in the step (1);
Wherein, the epidemic wood recognition model selects YOLOv object detection models;
(3) Performing epidemic detection and identification on the images of all the aerial image raw sheets by using the epidemic identification model trained in the step (2) to obtain all the epidemic and the position information of all the epidemic in the images of each aerial image raw sheet;
(4) Embedding special marks with obvious features at epidemic wood positions of the images of each aerial image original piece obtained in the step (3), and splicing all aerial image original pieces embedded with the special marks into an orthographic image with longitude and latitude coordinates by using image splicing software or an algorithm;
the special mark with obvious characteristics is a white solid circle with the radius of 20 pixels;
(5) Selecting an image of which part contains a special mark in the orthographic image as a special mark training sample, and marking the special mark in the special mark training sample image by an image marking tool;
(6) Selecting an object detection model based on a deep learning technology as a special mark recognition model, and training the special mark recognition model by using the special mark training sample generated in the step (5);
The special mark recognition model selects YOLOv models;
(7) Performing special mark recognition on all the orthographic images by using the special mark recognition model trained in the step (6) to obtain all the special marks and the position information thereof in the orthographic images;
(8) And (3) converting the coordinate information of the special mark generated in the step (7) into longitude and latitude coordinates, and acquiring the number of epidemic woods in the target mountain range and the longitude and latitude of each plant of epidemic woods, thereby realizing the identification and positioning of the plant of epidemic woods accurately.
2. The method for identifying and positioning pine wood nematode disease wood based on aerial photo of raw sheets of unmanned aerial vehicle according to claim 1, wherein the method for acquiring the disease wood training sample in step (1) comprises the steps of presetting a target mountain forest area range, planning and setting a route, setting a plurality of phase control points on the ground, shooting raw sheets of a mountain forest area by the unmanned aerial vehicle on a camera at fixed altitude at constant speed, uniformly cutting each raw sheet into 768 x 768 pixel small images every fixed time after the raw sheets of the mountain forest area are shot by the unmanned aerial vehicle every fixed time, randomly selecting partial small images cut out by the raw sheets, and manually selecting all the disease wood in the images by using an image marking tool labelImg to obtain the disease wood training sample.
3. The method for identifying and positioning pine wood nematode epidemic wood based on the unmanned aerial vehicle aerial photo original sheet according to claim 1, wherein YOLOv object detection model in the step (2) is used as a reference model for epidemic wood identification, the value of the last full-connection layer of the YOLOv object detection model is set to 2, the last full-connection layer is used for representing two epidemic wood types of new-born epidemic wood and past-year epidemic wood in the current year, and all epidemic wood training samples are represented by 4:1:1 is divided into a training set, a verification set and a test set, the loss function is calculated through the training set, the model parameters are updated through back propagation, the super parameters of the model are optimized through the verification set, and finally the test set is used for testing the identification accuracy of the model under the epidemic training sample data set.
4. The method for identifying and positioning pine wood nematode disease wood based on unmanned aerial vehicle aerial photo raw sheets is characterized in that the method for detecting and identifying the pine wood nematode disease wood based on unmanned aerial vehicle aerial photo raw sheets in the step (3) comprises the steps of cutting all aerial photo image raw sheets shot by an unmanned aerial vehicle, uniformly cutting each raw sheet into 768 x 768 pixel small images, inputting the small images into a trained epidemic wood identification model for carrying out epidemic wood identification, outputting the epidemic wood identification model into pixel positions, epidemic wood types and confidence degrees of epidemic wood identification frames in the small images, and finally converting the positions of all the epidemic wood identification frames belonging to the raw sheets into the positions of the epidemic wood identification frames in the raw sheets according to the positions of each small image in the raw sheets.
5. The method for identifying and positioning pine wood nematode disease wood based on unmanned aerial vehicle aerial photo raw sheets according to claim 1, wherein in the step (4), all raw sheets embedded with special marks are guided into image splicing software or algorithm for splicing, an orthographic image with longitude and latitude coordinates is generated, and each pixel point in the orthographic image corresponds to one longitude and latitude coordinate.
6. The method for identifying and positioning pine wood nematode disease wood based on aerial photo of unmanned aerial vehicle raw sheets according to claim 1, wherein the special mark training sample in the step (5) is obtained by segmenting an obtained orthographic image into 768 x 768 pixel small images, and randomly selecting a plurality of small images containing special marks as the special mark training sample, and the image marking tool is labelImg marking tools.
7. The method for identifying and positioning pine wood nematode disease wood based on the unmanned aerial vehicle aerial photo original sheet according to claim 1, wherein the value of the last full-connection layer of the special mark identification model is set to be 1.
8. The method for identifying and positioning pine wood nematode disease wood based on the unmanned aerial vehicle aerial photo original sheet according to claim 1, wherein the method for identifying the special mark of all the orthographic images in the step (7) comprises the steps of inputting the segmented orthographic image small images into a trained special mark identification model for identification, outputting the special mark identification model into the pixel positions of the special marks in the small images and the confidence coefficient, namely the pixel positions of the epidemic wood in each small image, setting the confidence coefficient threshold to be 0.8, and finally reserving the special mark identification result with the confidence coefficient exceeding the threshold.
9. The method for identifying and positioning pine wood nematode disease wood based on the unmanned aerial vehicle aerial photo original sheet according to claim 1, wherein the method for converting the pine wood nematode disease wood based on the unmanned aerial vehicle aerial photo original sheet into longitude and latitude coordinates in the step (8) comprises the steps of obtaining pixel positions of all special mark center points in an orthographic image according to pixel positions of small images in the orthographic image and pixel positions of special marks in each small image, and obtaining longitude and latitude coordinates of the center points through corresponding relations between the pixel positions and the longitude and latitude coordinates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110820396.4A CN113537089B (en) | 2021-07-20 | 2021-07-20 | Pine wood nematode epidemic wood identification and positioning method based on unmanned aerial vehicle aerial photography original sheet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110820396.4A CN113537089B (en) | 2021-07-20 | 2021-07-20 | Pine wood nematode epidemic wood identification and positioning method based on unmanned aerial vehicle aerial photography original sheet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113537089A CN113537089A (en) | 2021-10-22 |
| CN113537089B true CN113537089B (en) | 2024-07-26 |
Family
ID=78100488
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110820396.4A Active CN113537089B (en) | 2021-07-20 | 2021-07-20 | Pine wood nematode epidemic wood identification and positioning method based on unmanned aerial vehicle aerial photography original sheet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537089B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116359962A (en) * | 2021-12-20 | 2023-06-30 | 成都圭目机器人有限公司 | Estimated position correction method, apparatus, terminal and storage medium |
| CN115471478A (en) * | 2022-09-19 | 2022-12-13 | 安徽大学 | Unmanned aerial vehicle remote sensing monitoring method for pine tree diseases based on yolov5 |
| CN115619719B (en) * | 2022-09-26 | 2025-08-22 | 华南农业大学 | A method for detecting pine wood nematode diseased trees based on an improved Yolo v3 network model |
| CN116442662B (en) * | 2023-04-12 | 2025-06-20 | 生态环境部南京环境科学研究所 | A method for determining the spatial structure of biodiversity conservation areas |
| CN116912476B (en) * | 2023-07-05 | 2024-05-31 | 农芯(南京)智慧农业研究院有限公司 | Remote sensing monitoring rapid positioning method and related device for pine wood nematode disease unmanned aerial vehicle |
| CN120088687A (en) * | 2025-05-06 | 2025-06-03 | 东方电子股份有限公司 | A reflective film detection method based on aerial images and a reflective film detection system thereof |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108037770A (en) * | 2017-11-22 | 2018-05-15 | 国网山东省电力公司济宁供电公司 | Unmanned plane power transmission line polling system and method based on artificial intelligence |
| CN109766830A (en) * | 2019-01-09 | 2019-05-17 | 深圳市芯鹏智能信息有限公司 | A kind of ship seakeeping system and method based on artificial intelligence image procossing |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7042470B2 (en) * | 2001-03-05 | 2006-05-09 | Digimarc Corporation | Using embedded steganographic identifiers in segmented areas of geographic images and characteristics corresponding to imagery data derived from aerial platforms |
| JP4280572B2 (en) * | 2003-07-17 | 2009-06-17 | アジア航測株式会社 | Automatic orientation method using special marks |
| CN103366555B (en) * | 2013-07-01 | 2015-06-17 | 中国人民解放军第三军医大学第三附属医院 | Aerial image-based traffic accident scene image rapid generation method and system |
| CN108710875B (en) * | 2018-09-11 | 2019-01-08 | 湖南鲲鹏智汇无人机技术有限公司 | A kind of take photo by plane road vehicle method of counting and device based on deep learning |
| CN109948563A (en) * | 2019-03-22 | 2019-06-28 | 华南农业大学 | A deep learning-based detection and localization method for dead trees of pine wood nematode disease |
| CN110595476B (en) * | 2019-08-30 | 2021-07-06 | 天津航天中为数据系统科技有限公司 | Unmanned aerial vehicle landing navigation method and device based on GPS and image visual fusion |
| CN112001966B (en) * | 2020-08-03 | 2023-06-09 | 南京理工大学 | Positioning and tracking method of display screen in flight training AR system |
-
2021
- 2021-07-20 CN CN202110820396.4A patent/CN113537089B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108037770A (en) * | 2017-11-22 | 2018-05-15 | 国网山东省电力公司济宁供电公司 | Unmanned plane power transmission line polling system and method based on artificial intelligence |
| CN109766830A (en) * | 2019-01-09 | 2019-05-17 | 深圳市芯鹏智能信息有限公司 | A kind of ship seakeeping system and method based on artificial intelligence image procossing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113537089A (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113537089B (en) | Pine wood nematode epidemic wood identification and positioning method based on unmanned aerial vehicle aerial photography original sheet | |
| CN106403904B (en) | A kind of calculation method and system of the landscape scale vegetation coverage based on unmanned plane | |
| CN109117811B (en) | A system and method for estimating urban vegetation coverage based on low-altitude remote sensing measurement technology | |
| CN109815300B (en) | A vehicle positioning method | |
| JP6869264B2 (en) | Information processing equipment, information processing methods and programs | |
| CN106909886B (en) | A high-precision traffic sign detection method and system based on deep learning | |
| US11769225B2 (en) | Image processing apparatus, image processing method, and program | |
| CN113378754B (en) | A method for monitoring bare soil on a construction site | |
| CN106897681A (en) | A kind of remote sensing images comparative analysis method and system | |
| CN114140665A (en) | A Dense Small Object Detection Method Based on Improved YOLOv5 | |
| CN113255578A (en) | Traffic identification recognition method and device, electronic equipment and storage medium | |
| CN109813286A (en) | A kind of lodging disaster remote sensing damage identification method based on unmanned plane | |
| CN108416263A (en) | A kind of drone height measurement method of low cost suitable for the monitoring of agriculture feelings low-altitude remote sensing | |
| CN114140703A (en) | Intelligent recognition method and system for forest pine wood nematode diseases | |
| CN111552762A (en) | Orchard planting digital map management method and system based on fruit tree coding | |
| CN116486267A (en) | A large-area bamboo forest counting method based on UAV aerial images | |
| CN114140428A (en) | Method and system for detection and identification of larch caterpillar pests based on YOLOv5 | |
| CN118072160A (en) | A method and system for accurately estimating biomass at the scale of individual trees based on improved Mask R-CNN | |
| CN112163639B (en) | A crop lodging classification method based on height distribution feature vector | |
| CN111950524B (en) | Orchard local sparse mapping method and system based on binocular vision and RTK | |
| CN113112590A (en) | Method for acquiring ecological change and vegetation index in ecological water delivery engineering | |
| CN112613437A (en) | High-accuracy illegal building identification method | |
| CN118482698A (en) | Surveying method of surveying drone without field work | |
| CN117726955A (en) | A method and system for identifying termite indicators using optical and infrared drone data | |
| CN114998584B (en) | A method, device, system and storage medium for detecting abnormal vegetation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |