Ecological sample plot monitoring method and system based on computer vision

Abstract

The invention is suitable for the technical field of sample plot monitoring, and provides an ecological sample plot monitoring method and system based on computer vision, and the method comprises the following steps: collecting aerial images of an ecological sample plot, each aerial image corresponding to one piece of aerial parameter information; edge cutting is carried out on the aerial images according to the aerial parameter information, and all the aerial images after edge cutting are spliced to obtain an ecological sample plot image; performing image processing and image segmentation on the ecological sample plot image, segmenting the ecological sample plot image into a plurality of areas, and extracting key feature information of each area; identifying ecological information of each region according to the key feature information, and obtaining identification credibility; determining an area range needing to be recollected and collection coordinates based on the identification credibility; and re-determining the ecological information of each area according to the re-collected aerial images. According to the method, blind re-mining is avoided according to the identification credibility, and the accuracy and integrity of the ecological information are further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of sample plot monitoring, and in particular to an ecological sample plot monitoring method and system based on computer vision. Background Art

[0002] With the rapid development of remote sensing technology and computer vision technology, the use of drone aerial photography combined with image processing technology to monitor ecological plots has become an emerging and efficient method. However, ecological plots usually contain a variety of complex ecological and environmental elements, such as ecological types, vegetation types, etc. The performance characteristics of these elements in the image are different. How to accurately and efficiently segment the image and extract key feature information has become a difficult problem. In addition, due to the influence of factors such as lighting conditions and shooting angles, the images obtained by aerial photography may have incomplete or distorted information, affecting the accuracy of ecological information recognition. Therefore, it is necessary to provide an ecological plot monitoring method and system based on computer vision to solve the above problems. Summary of the invention

[0003] In view of the deficiencies in the prior art, the object of the present invention is to provide an ecological plot monitoring method and system based on computer vision to solve the problems existing in the above-mentioned background technology.

[0004] The present invention is implemented as follows: a computer vision-based ecological plot monitoring method, the method comprising the following steps: Collecting aerial images of the ecological plot, each aerial image corresponds to an aerial parameter information, and the aerial parameter information includes aerial coordinates, aerial altitude and shooting focal length; The aerial images are edge-cropped according to the aerial photography parameter information, and all the aerial images after edge-cropping are spliced ​​to obtain the ecological sample plot image; Perform image processing and image segmentation on the ecological sample plot image, divide the ecological sample plot image into several areas, and extract the key feature information of each area; Identify the ecological information of each area based on key feature information and obtain the identification credibility; Determine the area scope and collection coordinates that need to be recollected based on the recognition confidence; The secondary collected aerial images are received, and the ecological information of each area is re-determined based on the secondary collected aerial images.

[0005] As a further solution of the present invention, the step of performing edge cropping on the aerial image according to the aerial photography parameter information and splicing all the aerial images after edge cropping to obtain the ecological sample plot image specifically includes: The actual ground size L represented by each pixel in the aerial image is calculated based on the aerial photography parameter information. The actual ground size L=2×h×tanα×W, where h represents the aerial photography altitude, W represents the aerial image size, and α represents half of the camera's viewing angle, which is calculated based on the shooting focal length and sensor size. Determine the edge overlap area of ​​each aerial image according to the actual ground size and the aerial coordinates, and crop the edge overlap area; Image fusion and splicing are performed based on feature point matching and RANSAC algorithm to obtain ecological sample plot images.

[0006] As a further solution of the present invention: the steps of performing image processing and image segmentation on the ecological sample plot image, segmenting the ecological sample plot image into a plurality of regions, and extracting key feature information of each region specifically include: The ecological sample plot image was converted from RGB color space to HSV color space, and then filtered, denoised and contrast enhanced. The ecological sample plot image is preliminarily segmented based on the H component of the HSV color space, and further segmented according to the region growing algorithm; The texture features, color features and shape features of each area are extracted to obtain key feature information.

[0007] As a further solution of the present invention: the step of identifying the ecological information of each area according to the key feature information and obtaining the recognition credibility specifically includes: The fused feature vector is obtained by performing feature fusion according to texture features, color features and shape features, and the ecological label of the area is obtained based on the fused feature vector and the trained recognition model; Retrieving a reference set corresponding to the eco-label, and using cosine similarity to find K samples that are nearest neighbors to the fused feature vector in the reference set; The comprehensive similarity index is obtained according to the similarity between the K nearest neighbor samples and the fused feature vector, and the recognition credibility is determined according to the comprehensive similarity index.

[0008] As a further solution of the present invention: the step of determining the area range and acquisition coordinates that need to be re-collected based on the recognition credibility specifically includes: Based on the recognition confidence, the area that needs to be re-collected is determined, and the area range is obtained by summarizing; The regional range is covered based on the maximum secondary acquisition block, and the maximum secondary acquisition block has a constant coverage area; The center coordinates of each largest secondary acquisition block involved in the coverage are determined, and the center coordinates are acquisition coordinates.

[0009] As a further solution of the present invention: the steps of performing preliminary segmentation of the ecological sample plot image based on the H component of the HSV color space and further segmenting the ecological sample plot image according to the region growing algorithm specifically include: Traverse each pixel of the ecological sample plot image, determine the threshold range that the H component value of each pixel falls into, each threshold range corresponds to an ecological element, and perform preliminary segmentation based on the ecological element; Select seed points and determine similarity criteria; Starting from each seed point, check whether the pixels in its neighborhood meet the similarity criterion. If so, add the pixels to the current area; Repeat the previous step until no new pixels can be added to the current region and the seed point completes the region growth.

[0010] Another object of the present invention is to provide an ecological plot monitoring system based on computer vision, the system comprising: An aerial image acquisition module is used to collect aerial images of ecological plots, each aerial image corresponds to an aerial parameter information, and the aerial parameter information includes aerial coordinates, aerial altitude and shooting focal length; An image cropping and splicing module is used to crop the edges of the aerial images according to the aerial photography parameter information, and to splice all the aerial images after the edges have been cropped to obtain an ecological sample plot image; The key feature extraction module is used to process and segment the ecological sample plot image, segment the ecological sample plot image into several regions, and extract the key feature information of each region; Image region recognition module, used to identify the ecological information of each region based on key feature information and obtain the recognition credibility; A secondary collection determination module is used to determine the area range and collection coordinates that need to be re-collected based on the recognition credibility; The ecological information determination module is used to receive the secondary collected aerial images and re-determine the ecological information of each area based on the secondary collected aerial images.

[0011] As a further solution of the present invention: the image cropping and splicing module includes: The actual ground size unit is used to calculate the actual ground size L represented by each pixel in the aerial image according to the aerial photography parameter information. The actual ground size L=2×h×tanα×W, where h represents the aerial photography altitude, W represents the aerial image size, and α represents half of the camera viewing angle, which is calculated by the shooting focal length and the sensor size. The overlap region clipping unit is used to determine the edge overlap region of each aerial image according to the actual ground size and the aerial coordinates, and to clip the edge overlap region; The image fusion and stitching unit is used to perform image fusion and stitching based on feature point matching and RANSAC algorithm to obtain ecological sample plot images.

[0012] As a further solution of the present invention: the key feature extraction module includes: A color space conversion unit is used to convert the ecological sample plot image from RGB color space to HSV color space for filtering, denoising and contrast enhancement; A sample plot image segmentation unit is used to perform preliminary segmentation of the ecological sample plot image based on the H component of the HSV color space, and further segment the ecological sample plot image according to a region growing algorithm; The key feature extraction unit is used to extract the texture features, color features and shape features of each area to obtain key feature information.

[0013] As a further solution of the present invention: the image region recognition module includes: An eco-label recognition unit, used to obtain a fused feature vector by performing feature fusion according to texture features, color features and shape features, and obtain an eco-label of the region based on the fused feature vector and a trained recognition model; A nearest neighbor sample search unit is used to retrieve a reference set corresponding to the eco-label, and use cosine similarity to search for K samples that are nearest neighbors to the fused feature vector in the reference set; The recognition credibility unit is used to obtain a comprehensive similarity index according to the similarity between the K nearest neighbor samples and the fused feature vector, and determine the recognition credibility according to the comprehensive similarity index.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention performs edge cropping on aerial images according to aerial photography parameter information, removes overlapping and blurred parts in the image, reduces the amount of calculation and interference in subsequent image processing, and improves processing efficiency and accuracy. The ecological information of each area is identified according to key feature information, and the recognition credibility is obtained at the same time. According to the recognition credibility, the area range and acquisition coordinates that need to be re-collected are automatically determined, and the coordinates of the area for secondary collection are accurately determined, ensuring that the final image information obtained is complete and accurate, and the feature recognition results are more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flowchart of an ecological plot monitoring method based on computer vision.

[0016] Figure 2 The flowchart of cropping aerial images in an ecological plot monitoring method based on computer vision.

[0017] Figure 3The figure is a flowchart of image processing in an ecological plot monitoring method based on computer vision.

[0018] Figure 4 Flowchart for identifying ecological information in an ecological plot monitoring method based on computer vision.

[0019] Figure 5 This is a flowchart for determining the area scope and collection coordinates for re-collection in an ecological sample plot monitoring method based on computer vision.

[0020] Figure 6 The figure is a flowchart of ecological plot image segmentation in an ecological plot monitoring method based on computer vision.

[0021] Figure 7 This is a structural diagram of an ecological plot monitoring system based on computer vision. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0023] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0024] like Figure 1 As shown, an embodiment of the present invention provides an ecological plot monitoring method based on computer vision, the method comprising the following steps: S100, collecting aerial images of the ecological plot, each aerial image corresponding to an aerial parameter information, the aerial parameter information including aerial coordinates, aerial altitude and shooting focal length; S200, performing edge cropping on the aerial image according to the aerial photography parameter information, and splicing all the aerial images after edge cropping to obtain an ecological sample plot image; S300, performing image processing and image segmentation on the ecological sample plot image, segmenting the ecological sample plot image into a plurality of regions, and extracting key feature information of each region; S400, identifying ecological information of each area according to key feature information, and obtaining identification credibility; S500, determining the area range and acquisition coordinates that need to be re-collected based on the recognition credibility; S600, receiving the secondary collected aerial images, and re-determining the ecological information of each area according to the secondary collected aerial images.

[0025] In the embodiment of the present invention, when collecting aerial images of ecological plots, an aerial parameter information is automatically set for each aerial image, and the aerial parameter information includes aerial coordinates, aerial altitude, and shooting focal length. The drone is equipped with a positioning system and an altitude sensor, which can obtain coordinates and altitude in real time. Then, the embodiment of the present invention will perform edge cropping on the aerial image according to the aerial parameter information, and splice all the aerial images after edge cropping to obtain the ecological plot image, so that the fusion and splicing effect is better. Then, the ecological plot image will be processed and segmented, and the ecological plot image will be segmented into several areas. Each area belongs to the same ecological type, which includes grassland, forest, water body, farmland, etc. The key feature information of each area is extracted, and then the ecological information of each area is identified based on the key feature information. The ecological information is a subdivision of the ecological type. For example, the ecological information of the forest ecology is a specific vegetation type. At the same time, the recognition credibility is obtained. The recognition credibility reflects the accuracy of the ecological information obtained by identification. If the recognition credibility is low, it means that the recognition result is not accurate enough. The embodiment of the present invention will determine the area range and acquisition coordinates that need to be re-collected according to all the recognition credibility, and perform secondary aerial photography based on the area range and acquisition coordinates to obtain the secondary collected aerial images. The ecological information of each area is re-determined based on the secondary collected aerial images. In this way, it is ensured that the image information finally obtained is complete and accurate, and the feature recognition result is credible.

[0026] like Figure 2 As shown, as a preferred embodiment of the present invention, the step of performing edge cropping on the aerial image according to the aerial photography parameter information and stitching all the edge cropped aerial images to obtain the ecological plot image specifically includes: S201, calculating the actual ground size L represented by each pixel in the aerial image according to the aerial photography parameter information, the actual ground size L=2×h×tanα×W, where h represents the aerial photography height, W represents the aerial image size, and α represents half of the camera viewing angle, which is calculated by the shooting focal length and the sensor size; S202, determining the edge overlap region of each aerial image according to the actual ground size and the aerial coordinates, and cropping the edge overlap region; S203, performing image fusion and splicing based on feature point matching and RANSAC algorithm to obtain an ecological sample plot image.

[0027] In the embodiment of the present invention, in order to perform accurate cropping and splicing, it is first necessary to calculate the actual ground size L represented by each pixel in the aerial image according to the aerial parameter information, and the actual ground size L=2×h×tanα×W, and then determine the edge overlap area of ​​each aerial image according to the actual ground size and the aerial coordinates, and crop the edge overlap area. The aerial coordinates are the center point of each aerial image. Cropping the edge area can remove the overlapped part on the one hand, and remove the part of the image with blurred edges on the other hand. Finally, the image fusion and splicing will be performed according to the feature point matching (such as SIFT, SURF) and RANSAC algorithm to obtain the ecological sample site image.

[0028] like Figure 3 As shown, as a preferred embodiment of the present invention, the steps of performing image processing and image segmentation on the ecological sample plot image, segmenting the ecological sample plot image into several regions, and extracting key feature information of each region specifically include: S301, converting the ecological sample plot image from RGB color space to HSV color space, performing filtering denoising and contrast enhancement; S302, performing preliminary segmentation on the ecological sample plot image based on the H component of the HSV color space, and further segmenting the ecological sample plot image according to a region growing algorithm; S303, extracting texture features, color features, and shape features of each region to obtain key feature information.

[0029] In the embodiment of the present invention, the ecological sample image is converted from the RGB color space to the HSV color space. The H (hue) component in the HSV color space is particularly useful for distinguishing different ecological types. Then, Gaussian filtering, median filtering and other algorithms are applied to remove noise in the image and enhance contrast. Then, the ecological sample image is preliminarily segmented based on the H component of the HSV color space, and the ecological sample image is further segmented according to the regional growing algorithm to obtain several regions. Finally, the texture features, color features and shape features of each region are extracted to obtain key feature information, wherein the texture features are extracted using gray level co-occurrence matrix (GLCM), local binary pattern (LBP) and other methods; the color features are obtained by calculating the average color of each region, color histogram, etc.

[0030] like Figure 4 As shown, as a preferred embodiment of the present invention, the step of identifying the ecological information of each area according to the key feature information and obtaining the recognition credibility specifically includes: S401, performing feature fusion according to texture features, color features and shape features to obtain a fused feature vector, and obtaining an ecological label of the region based on the fused feature vector and a trained recognition model; S402, retrieve a reference set corresponding to the eco-label, and use cosine similarity to find K samples that are nearest neighbors to the fused feature vector in the reference set; S403, obtaining a comprehensive similarity index according to the similarity between the K nearest neighbor samples and the fused feature vector, and determining the recognition credibility according to the comprehensive similarity index.

[0031] In the embodiment of the present invention, a recognition model is trained in advance. Specifically, a decision tree, random forest, support vector machine (SVM), neural network and other models are selected for training. Regional data (i.e., labeled data) with known ecological information is used to train the recognition model. During or after the training, a validation set or a test set is used to evaluate the performance of the model. When used, a fusion feature vector is obtained by feature fusion according to texture features, color features and shape features, and an ecological label of the region is obtained based on the fusion feature vector and the trained recognition model; a reference set corresponding to the ecological label is retrieved, and the training set used in training and the validation set used in verification are both reference sets. In the reference set, cosine similarity is used to find K samples that are the nearest neighbors of the fusion feature vector, where K is a preset constant value, and then a comprehensive similarity index is obtained based on the similarity between the K nearest neighbor samples and the fusion feature vector. The comprehensive similarity index can be the average similarity of the K samples, and finally the recognition credibility is determined based on the comprehensive similarity index, and the recognition credibility is proportional to the comprehensive similarity index.

[0032] like Figure 5 As shown, as a preferred embodiment of the present invention, the step of determining the area range and acquisition coordinates that need to be re-collected based on the recognition credibility specifically includes: S501, determining the area that needs to be re-collected based on the recognition credibility, and summarizing to obtain the area range; S502, covering the area range based on the maximum secondary acquisition block, where the maximum secondary acquisition block has a constant coverage area; S503: Determine the center coordinates of each largest secondary acquisition block involved in coverage, where the center coordinates are acquisition coordinates.

[0033] In an embodiment of the present invention, the area that needs to be recaptured will be determined based on the recognition credibility. For example, when the recognition credibility is lower than the credibility threshold, the corresponding regional image needs to be recaptured. These areas are summarized to obtain the regional range, and then the regional range is covered based on the maximum secondary acquisition block. The maximum secondary acquisition block has a constant coverage area. The maximum secondary acquisition block is a block size obtained using the set aerial photography height and shooting focal length, which can clearly capture the content in the block. The principle of covering the regional range is: use the minimum maximum secondary acquisition blocks as much as possible to cover the regional range in all directions, and finally determine the center coordinates of each maximum secondary acquisition block involved in the coverage, and the center coordinates are the acquisition coordinates.

[0034] like Figure 6 As shown, as a preferred embodiment of the present invention, the steps of performing preliminary segmentation of the ecological sample plot image based on the H component of the HSV color space and further segmenting the ecological sample plot image according to the region growing algorithm specifically include: S3021, traversing each pixel of the ecological sample plot image, determining a threshold range within which the H component value of each pixel falls, each threshold range corresponding to an ecological element, and performing preliminary segmentation based on the ecological element; S3022, selecting a seed point and determining a similarity criterion; S3023, starting from each seed point, checking whether the pixels in its neighborhood meet the similarity criterion, and if so, adding the pixels to the current area; S3024, repeat the previous step until no new pixels can be added to the current region and the seed point completes the region growth.

[0035] In an embodiment of the present invention, when performing image segmentation, each pixel of the ecological sample plot image is first traversed to determine the threshold range within which the H component value of each pixel falls. Each threshold range corresponds to an ecological element, and the ecological element is grassland, forest, water body, farmland, etc., and then preliminary segmentation is performed based on the type of ecological element. Next, it is necessary to select seed points. The seed points can be randomly selected by setting the seed density, and then the similarity criterion is determined. The similarity criterion can be determined based on the H component, the S component, and the V component. Starting from each seed point, check whether the pixels in its neighborhood meet the similarity criterion. If so, the pixel is added to the current area, and this step is repeated until no new pixels can be added to the current area, and the seed point completes the regional growth. Finally, it is necessary to merge adjacent similar areas or remove too small noise areas.

[0036] like Figure 7 As shown, an embodiment of the present invention further provides an ecological plot monitoring system based on computer vision, the system comprising: The aerial image acquisition module 100 is used to collect aerial images of the ecological plot, each aerial image corresponds to an aerial parameter information, and the aerial parameter information includes aerial coordinates, aerial altitude and shooting focal length; The image cropping and splicing module 200 is used to crop the edges of the aerial images according to the aerial photography parameter information, and to splice all the aerial images after the edges have been cropped to obtain an ecological sample plot image; The key feature extraction module 300 is used to perform image processing and image segmentation on the ecological sample plot image, segment the ecological sample plot image into a plurality of regions, and extract key feature information of each region; An image region identification module 400 is used to identify ecological information of each region based on key feature information and obtain identification credibility; A secondary collection determination module 500 is used to determine the area range and collection coordinates that need to be re-collected based on the recognition credibility; The ecological information determination module 600 is used to receive the secondary collected aerial images and re-determine the ecological information of each area according to the secondary collected aerial images.

[0037] As a preferred embodiment of the present invention, the image cropping and splicing module 200 includes: The actual ground size unit is used to calculate the actual ground size L represented by each pixel in the aerial image according to the aerial photography parameter information. The actual ground size L=2×h×tanα×W, where h represents the aerial photography altitude, W represents the aerial image size, and α represents half of the camera viewing angle, which is calculated by the shooting focal length and the sensor size. The overlap region clipping unit is used to determine the edge overlap region of each aerial image according to the actual ground size and the aerial coordinates, and to clip the edge overlap region; The image fusion and stitching unit is used to perform image fusion and stitching based on feature point matching and RANSAC algorithm to obtain ecological sample plot images.

[0038] As a preferred embodiment of the present invention, the key feature extraction module 300 includes: A color space conversion unit is used to convert the ecological sample plot image from RGB color space to HSV color space for filtering, denoising and contrast enhancement; A sample plot image segmentation unit is used to perform preliminary segmentation of the ecological sample plot image based on the H component of the HSV color space, and further segment the ecological sample plot image according to a region growing algorithm; The key feature extraction unit is used to extract the texture features, color features and shape features of each area to obtain key feature information.

[0039] As a preferred embodiment of the present invention, the image region recognition module 400 includes: An eco-label recognition unit, used to obtain a fused feature vector by performing feature fusion according to texture features, color features and shape features, and obtain an eco-label of the region based on the fused feature vector and a trained recognition model; A nearest neighbor sample search unit is used to retrieve a reference set corresponding to the eco-label, and use cosine similarity to search for K samples that are nearest neighbors to the fused feature vector in the reference set; The recognition credibility unit is used to obtain a comprehensive similarity index according to the similarity between the K nearest neighbor samples and the fused feature vector, and determine the recognition credibility according to the comprehensive similarity index.

[0040] The above only describes in detail the preferred embodiments of the present invention, which is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0041] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0042] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0043] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the disclosure in the specification and examples. This application is intended to cover any variations, uses or adaptations of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

Claims

1. A computer vision-based ecological plot monitoring method, characterized in that: The method comprises the following steps: Collecting aerial images of the ecological sample plot, each aerial image corresponds to an aerial parameter information, and the aerial parameter information includes aerial coordinates, aerial altitude and shooting focal length; The aerial images are edge-cropped according to the aerial photography parameter information, and all the aerial images after edge-cropping are spliced ​​to obtain the ecological sample plot image; Perform image processing and image segmentation on the ecological sample plot image, divide the ecological sample plot image into several areas, and extract the key feature information of each area; Identify the ecological information of each area based on key feature information and obtain the identification credibility; Determine the area scope and collection coordinates that need to be recollected based on the recognition confidence; The secondary collected aerial images are received, and the ecological information of each area is re-determined based on the secondary collected aerial images.

2. The ecological plot monitoring method based on computer vision according to claim 1, characterized in that: The step of performing edge cropping on the aerial image according to the aerial photography parameter information and splicing all the edge cropped aerial images to obtain the ecological sample plot image specifically includes: The actual ground size L represented by each pixel in the aerial image is calculated based on the aerial photography parameter information. The actual ground size L=2×h×tanα×W, where h represents the aerial photography altitude, W represents the aerial image size, and α represents half of the camera's viewing angle, which is calculated based on the shooting focal length and sensor size. Determine the edge overlap area of ​​each aerial image according to the actual ground size and the aerial coordinates, and crop the edge overlap area; Image fusion and splicing are performed based on feature point matching and RANSAC algorithm to obtain ecological sample plot images.

3. The ecological plot monitoring method based on computer vision according to claim 1, characterized in that: The steps of performing image processing and image segmentation on the ecological sample plot image, segmenting the ecological sample plot image into a plurality of regions, and extracting key feature information of each region specifically include: The ecological sample plot image was converted from RGB color space to HSV color space, and then filtered, denoised and contrast enhanced. The ecological sample plot image is preliminarily segmented based on the H component of the HSV color space, and further segmented according to the region growing algorithm; The texture features, color features and shape features of each area are extracted to obtain key feature information.

4. The method for ecological plot monitoring based on computer vision according to claim 1, characterized in that: The step of identifying the ecological information of each area according to the key feature information and obtaining the recognition credibility specifically includes: The fused feature vector is obtained by performing feature fusion according to texture features, color features and shape features, and the ecological label of the area is obtained based on the fused feature vector and the trained recognition model; Retrieving a reference set corresponding to the eco-label, and using cosine similarity to find K samples that are nearest neighbors to the fused feature vector in the reference set; The comprehensive similarity index is obtained according to the similarity between the K nearest neighbor samples and the fused feature vector, and the recognition credibility is determined according to the comprehensive similarity index.

5. The method for ecological plot monitoring based on computer vision according to claim 1, characterized in that: The step of determining the area range and acquisition coordinates that need to be re-collected based on the recognition credibility specifically includes: Based on the recognition confidence, the area that needs to be re-collected is determined, and the area range is obtained by summarizing; The regional range is covered based on the maximum secondary acquisition block, and the maximum secondary acquisition block has a constant coverage area; The center coordinates of each largest secondary acquisition block involved in the coverage are determined, and the center coordinates are acquisition coordinates.

6. The method for ecological plot monitoring based on computer vision according to claim 3, characterized in that: The steps of preliminarily segmenting the ecological sample plot image based on the H component of the HSV color space and further segmenting the ecological sample plot image according to the region growing algorithm specifically include: Traverse each pixel of the ecological sample plot image, determine the threshold range that the H component value of each pixel falls into, each threshold range corresponds to an ecological element, and perform preliminary segmentation based on the ecological element; Select seed points and determine similarity criteria; Starting from each seed point, check whether the pixels in its neighborhood meet the similarity criterion. If so, add the pixels to the current area; Repeat the previous step until no new pixels can be added to the current region and the seed point completes the region growth.

7. An ecological plot monitoring system based on computer vision, characterized in that: The system comprises: An aerial image acquisition module is used to collect aerial images of ecological plots, each aerial image corresponds to an aerial parameter information, and the aerial parameter information includes aerial coordinates, aerial altitude and shooting focal length; An image cropping and splicing module is used to crop the edges of the aerial images according to the aerial photography parameter information, and to splice all the aerial images after the edges have been cropped to obtain an ecological sample plot image; The key feature extraction module is used to process and segment the ecological sample plot image, segment the ecological sample plot image into several regions, and extract the key feature information of each region; Image region recognition module, used to identify the ecological information of each region based on key feature information and obtain the recognition credibility; A secondary collection determination module is used to determine the area range and collection coordinates that need to be re-collected based on the recognition credibility; The ecological information determination module is used to receive the secondary collected aerial images and re-determine the ecological information of each area based on the secondary collected aerial images.

8. The computer vision-based ecological plot monitoring system according to claim 7, characterized in that: The image cropping and splicing module includes: The actual ground size unit is used to calculate the actual ground size L represented by each pixel in the aerial image according to the aerial photography parameter information. The actual ground size L=2×h×tanα×W, where h represents the aerial photography altitude, W represents the aerial image size, and α represents half of the camera viewing angle, which is calculated by the shooting focal length and the sensor size. The overlap region clipping unit is used to determine the edge overlap region of each aerial image according to the actual ground size and the aerial coordinates, and to clip the edge overlap region; The image fusion and stitching unit is used to perform image fusion and stitching based on feature point matching and RANSAC algorithm to obtain ecological sample plot images.

9. The computer vision-based ecological plot monitoring system according to claim 7, characterized in that: The key feature extraction module comprises: A color space conversion unit is used to convert the ecological sample plot image from RGB color space to HSV color space for filtering, denoising and contrast enhancement; A sample plot image segmentation unit is used to perform preliminary segmentation of the ecological sample plot image based on the H component of the HSV color space, and further segment the ecological sample plot image according to a region growing algorithm; The key feature extraction unit is used to extract the texture features, color features and shape features of each area to obtain key feature information.

10. The computer vision-based ecological plot monitoring system according to claim 7, characterized in that: The image region recognition module comprises: An eco-label recognition unit, used to obtain a fused feature vector by performing feature fusion according to texture features, color features and shape features, and obtain an eco-label of the region based on the fused feature vector and a trained recognition model; A nearest neighbor sample search unit is used to retrieve a reference set corresponding to the eco-label, and use cosine similarity to search for K samples that are nearest neighbors to the fused feature vector in the reference set; The recognition credibility unit is used to obtain a comprehensive similarity index according to the similarity between the K nearest neighbor samples and the fused feature vector, and determine the recognition credibility according to the comprehensive similarity index.

Images