An unmanned aerial vehicle-based investigation method for concentrated distribution of water birds around a nuclear power plant
By combining drone-based rasterization and flight path planning with deep learning-based target detection and species classification, the problem of identification and counting errors in waterbird distribution surveys around nuclear power plants has been solved, achieving efficient and accurate waterbird distribution surveys.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA INST FOR RADIATION PROTECTION
- Filing Date
- 2026-01-30
- Publication Date
- 2026-06-09
AI Technical Summary
Traditional methods such as zonal direct counting and cluster estimation are difficult to accurately identify and count species in waterbird distribution surveys around nuclear power plants, resulting in large errors in observation results, especially for waterbirds that are distributed in long-distance clusters.
A drone-based survey method was adopted, which involved gridding the survey area and planning a zigzag reciprocating flight path. Full coverage photography was achieved by combining horizontal and vertical overlap control. Target detection and classification networks were used to identify and count waterbirds. Image registration and deduplication mechanisms were combined, and finally, the counting results were corrected by manual review through sampling.
It enables efficient and accurate surveys of waterbird distribution in the vicinity of nuclear power plants under safe and low-interference conditions, reducing the risk of missed photos, improving the accuracy and stability of waterbird identification and counting, and making the output statistical results more reliable.
Smart Images
Figure CN122176565A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of wildlife surveys, specifically to a method for surveying the concentrated distribution of waterbirds around nuclear power plants based on unmanned aerial vehicles (UAVs). Background Technology
[0002] The zonal direct counting method and the cluster estimation method are the most commonly used survey methods for flocked waterbirds. The "Technical Specifications for Survey of Terrestrial Wild Animals and Their Habitats Part 4: Birds" (GB / T 37364.4—2024) and the "Technical Guidelines for Biodiversity Observation: Birds" (HJ 710.4—2014) provide detailed explanations of the zonal direct counting method and the cluster estimation method.
[0003] Wild birds are highly wary, and their habitats are far from areas of human activity. In particular, flocks of waterbirds often stay far from the water's edge, sometimes as far as 500 meters or even several kilometers away. Traditional methods such as zonal counting and cluster estimation rely on observation equipment such as long-zoom digital cameras and telescopes to observe and count birds near the water's edge. However, due to the great distance, it is often difficult to accurately identify species and make accurate counts, resulting in significant errors in the observation results. Summary of the Invention
[0004] To achieve the above and other related objectives, this invention discloses a method for surveying the concentrated distribution of waterbirds around nuclear power plants based on unmanned aerial vehicles (UAVs), comprising: The survey area was investigated to determine the drone launch points, flight safety constraints, and filming times. Obtain the boundaries of the survey area and rasterize it to divide the survey area into... Each grid, based on horizontal overlap. Vertical overlap Plan a zigzag reciprocating route and determine the heading step length. With the triggering strategy, the coverage of the survey area is not lower than the preset threshold; The drone performs flight photography according to the described route, and acquires image data with panoramic focal length and telephoto focal length respectively based on the same waypoint sequence, and records the pose information or timestamp information of each image; The image data is preprocessed, and the panoramic image and telephoto image are registered, mapping the telephoto image to the panoramic coordinate system. The registered images are used to detect waterbirds using a target detection network and candidate targets are filtered out according to a confidence threshold. The retained targets are then classified into species using a classification network. In adjacent images with overlapping coverage, duplicate targets of the same type are removed to obtain the machine counting results for each grid. A preset proportion of grid cells are extracted from all grid cells for verification, the error coefficient is calculated, the machine count is corrected, and the results of the number of each species and its spatial distribution are output.
[0005] Preferably, the horizontal scale of the rasterization is: Where W is the effective width of a single image on the ground. and satisfy: in, For flight altitude, The horizontal field of view of the camera. This represents the percentage of invalid edges.
[0006] Preferably, the heading step size Ground speed from heading Interval between shooting It is certain, and satisfies: and Further satisfy: in, For vertical overlap, This represents the effective coverage length of the image in the heading direction.
[0007] Preferably, the coverage rate not being lower than a preset threshold includes: The proportion of the intersection area of the boundary grid with the boundary polygon of the survey area to the grid area is calculated. When the proportion is less than a preset proportion threshold, a boundary reshooting route is generated to compensate for the boundary grid.
[0008] Preferably, the aerial photography satisfies low interference constraints, including: When the target bird species is a gull or a shorebird, the flight altitude should not be lower than [missing information]. When the target birds are geese, ducks, or cranes, the flight altitude should not be lower than [missing information]. ; and on the same waypoint sequence, respectively adopt Panoramic focal length and to Two acquisitions are performed at the telephoto focal length to obtain a panoramic image and a telephoto image.
[0009] Preferably, the deduplication includes: Map the center point of the detection box of the candidate target in the image to a panoramic coordinate system or a geographic coordinate system. Calculate the pairwise distance between targets of the same category in adjacent overlapping regions. When the distance is less than the deduplication threshold... If the targets are identified as the same target, the target with higher confidence is retained to generate a deduplicated target set.
[0010] Preferably, the error coefficient is based on species. Calculate separately, and satisfy: in, For manually reviewed raster sets, For grid Inner species The number of manual reviews For grid Inner species The number of machines counted. For species The total number of machine counts, For the corrected species total.
[0011] Preferably, the target detection network outputs confidence scores for candidate targets and applies a confidence threshold. Perform screening, and when the confidence level of the candidate target is less than The method further includes: discarding the candidate target when it is in time; the method also includes: performing hard example mining on samples with low confidence but high frequency of occurrence, and feeding them back into the training set after manual confirmation, so as to iteratively update the target detection network and / or the classification network.
[0012] Secondly, this invention discloses a system for surveying the concentrated distribution of waterbirds around nuclear power plants based on unmanned aerial vehicles (UAVs), comprising: The drone data acquisition unit is used to fly along the flight path and acquire panoramic and telephoto images, while recording pose information or timestamp information. The route planning unit is used to obtain and rasterize the boundaries of the survey area, based on the lateral overlap. Vertical overlap Generate a zigzag reciprocating flight path and determine the heading step size. ; The registration and embedding unit is used to register the panoramic image and the telephoto image and map the telephoto image to the panoramic coordinate system; The identification and counting unit includes a target detection subunit and a classification subunit. The target detection subunit is used to detect waterbird targets and filter them out according to a confidence threshold, and the classification subunit is used to output the species category. The identification and counting unit is also used to perform deduplication of the same type of target between adjacent overlapping images and to perform machine counting by raster. The sampling correction unit is used to extract a preset proportion of grid cells for verification and calculate the error coefficient to correct the machine count. The output unit is used to output the number and spatial distribution of each species. The units work together to implement the method.
[0013] Thirdly, the present invention discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method.
[0014] By adopting the above technical solutions, under the premise of meeting the safety and low-interference constraints around nuclear power plants, full-coverage photography can be achieved by gridding the survey area and planning a zigzag reciprocating flight path, combined with the control of lateral and longitudinal overlap. Automatic compensation is provided when boundary coverage is insufficient to reduce the risk of missed shots. Furthermore, the use of dual-focal-length acquisition along the same flight path enables the coordinate correspondence between panoramic distribution acquisition and long-focus individual feature acquisition. Combined with image registration and embedding, deep learning target detection, and species classification, the accuracy and applicability of waterbird identification and counting are improved. At the same time, a deduplication mechanism is introduced for overlapping areas, and the machine counting is corrected by combining sampling and manual verification to calculate the error coefficient, making the statistical results more stable, traceable, and easy to control in terms of quality. This enables a high-efficiency survey and output of results on the concentrated distribution of waterbirds around nuclear power plants. Attached Figure Description
[0015] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the scope of this disclosure. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein: Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation
[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0017] Reference Figure 1 This invention provides a method for surveying the concentrated distribution of waterbirds around nuclear power plants based on unmanned aerial vehicles (UAVs), comprising the following steps: The survey area was investigated to determine the drone launch points, flight safety constraints, and filming times. Obtain the boundaries of the survey area and rasterize it to divide the survey area into... Each grid, based on horizontal overlap. Vertical overlap Plan a zigzag reciprocating route and determine the heading step length. With the triggering strategy, the coverage of the survey area is not lower than the preset threshold; The drone performs flight photography according to the described route, and acquires image data with panoramic focal length and telephoto focal length respectively based on the same waypoint sequence, and records the pose information or timestamp information of each image; The image data is preprocessed, and the panoramic image and telephoto image are registered, mapping the telephoto image to the panoramic coordinate system. The registered images are used to detect waterbirds using a target detection network and candidate targets are filtered out according to a confidence threshold. The retained targets are then classified into species using a classification network. In adjacent images with overlapping coverage, duplicate targets of the same type are removed to obtain the machine counting results for each grid. A preset proportion of grid cells are extracted from all grid cells for verification, the error coefficient is calculated, the machine count is corrected, and the results of the number of each species and its spatial distribution are output.
[0018] Preferably, before the survey begins, the survey area is first determined, and the boundary polygon of the survey area is imported into the geographic information system. The boundary coordinates can be represented by WGS84 latitude and longitude. A task number and task metadata table are also established. The task metadata table includes at least: survey date, launch point coordinates, target group, planned number of flights, equipment model, camera focal length setting, and battery number.
[0019] To ensure usability for subsequent target detection and species classification, a minimum recognizable pixel scale constraint is pre-determined, such as requiring the side length of the bounding box for a single waterbird in a telephoto image to be no less than a preset pixel threshold; based on the relationship between the UAV's flight altitude and the camera's field of view, a preliminary recommended flight altitude is given. With telephoto magnification range (e.g.) to () is used for subsequent execution.
[0020] Preferably, the survey conducted in the investigation area includes: Select an open, unobstructed release point that meets takeoff and landing conditions; record wind speed, lighting conditions, and perimeter no-fly / restricted-fly zones, and set up return and alternate landing points; confirm the solar altitude angle and shooting direction on site, preferably shooting with front lighting or side-front lighting, and record the start and end times of the mission and lighting conditions for subsequent image preprocessing and quality assessment; follow the principle of "minimal disturbance" during the survey, and if significant signs of startled flight or migration are observed in the flock, immediately increase the flight altitude or terminate the flight, and record the disturbance event in the mission log for later review.
[0021] Preferably, in the preferred embodiment of the present invention, the launch location of the drone is selected based on the terrain, light direction, and wind intensity of the survey site. The preferred launch location is a flat area with a slope of ≤15°, free from tall obstacles, and maintains a sufficient safe distance from waterbird habitats; the solar altitude angle during the shooting period is between 30° and 60°; the angle between the drone's flight direction and the incident light direction is between 0° and 45° (front lighting or side-front lighting) to avoid overexposure or underexposure caused by backlighting; and the wind force during flight is ≤3 to reduce the risk of loss of control.
[0022] Preferably, route planning and gridding specifically includes: Convert the boundary polygon of the survey area from latitude and longitude coordinates to plane rectangular coordinates to calculate distance and area; a unified projection (such as UTM zone projection) can be used, and the projection parameters are recorded for result rotation.
[0023] Set horizontal overlap Vertical overlap ;in, The value range is 15% to 20%. The value range is 20% to 25%, and the horizontal scale of the raster in the rasterization is... Where W is the effective width of a single image on the ground. and satisfy: in, For flight altitude, The horizontal field of view of the camera. This is the edge invalidation ratio, used to remove edge areas with dark corners, distortion, or poor stability.
[0024] Subsequently, the survey area was divided into grids to form... Each grid cell is assigned a unique number.
[0025] In timed trigger scenarios, the speed is determined by the heading and ground speed. Interval between shooting Determine the heading step size : And based on the longitudinal overlap constraint : in, This represents the effective coverage length of the image in the heading direction.
[0026] In a fixed-distance trigger scenario, it can be Set to a fixed value and according to The trigger distance is calculated in reverse to improve overlap stability.
[0027] Flight routes are generated in a zigzag pattern, with the spacing between adjacent flight strips determined by... and Determine; generate waypoint sequences for each flight strip, and save the waypoint sequence order, heading, speed, triggering strategy, and estimated flight time in the mission file.
[0028] Preferably, in this embodiment of the invention, the starting point is set as the top-left vertex of the grid matrix. ,along Flying in the positive direction of the axis, after completing the first row of grid photography, along... Move in the positive direction of the axis Distance, turning along The second row of grids is photographed by flying in the negative direction of the axis; this process is repeated until all grids are photographed, forming a zigzag reciprocating flight path.
[0029] The proportion of the area of intersection between the boundary grid and the polygon of the survey area is calculated to the area of the grid. When the proportion is less than a preset threshold, a supplementary shooting route is automatically generated, such as a supplementary shooting trajectory that expands outward along the boundary or travels along the boundary, to ensure that the coverage meets the requirements. In this embodiment, the preset threshold is preferably 80%.
[0030] When the estimated flight time or battery power of a single sortie is insufficient to cover the entire grid, the survey area is divided into several sub-areas and the sorties are carried out separately. Each sub-area retains an independent mission number and records the sortie sequence for subsequent splicing and tracing.
[0031] Preferably, embodiments of the present invention use a drone equipped with a zoom lens, such as the DJI Movic 3 or DJI Ir 3S, and replace the propellers with silent ones to reduce noise interference. When the target birds are gulls or shorebirds, the flight altitude is not lower than [missing information]. When the target birds are geese, ducks, or cranes and storks, the flight altitude should not be lower than [missing information]. .
[0032] The drone loads the flight path task and performs the first data collection, using... A panoramic image is acquired using a panoramic focal length; a second acquisition is performed without changing the waypoint sequence. to Long focal lengths are used to acquire long focal length images to ensure that the two types of images correspond spatially.
[0033] Record pose information or timestamp information for each image; optionally, write waypoint number, exposure time, flight altitude, ground speed, etc. into the image EXIF or the accompanying mission log file for subsequent registration, alignment and quality control.
[0034] If a flock of birds is clearly startled and takes flight during the flight, strategies such as increasing altitude, increasing horizontal distance, reducing noise mode, or terminating the flight can be implemented, and the disturbance event should be associated with the corresponding image number and recorded.
[0035] Preferably, image preprocessing and registration specifically include: Preprocessing such as denoising, white balance correction, contrast enhancement, and sharpening is performed on the acquired images to improve the stability of target detection and classification.
[0036] Based on pose information or timestamps, each telephoto image is associated with its corresponding panoramic image and corresponding grid number to form a “panoramic-telephoto” pairing index.
[0037] Within the same grid area, feature point matching is performed on the panoramic image and the telephoto image. The geometric mapping from the telephoto image coordinate system to the panoramic image coordinate system (which can be a homography matrix or a block affine transformation) is calculated, and the telephoto image is mapped to the panoramic coordinate system, thus forming a registration result of telephoto image embedded in panoramic image.
[0038] When the registration residual is greater than the preset threshold or the number of matching points is insufficient, the corresponding image will be marked as a low-quality sample, and a manual review or re-acquisition suggestion will be triggered.
[0039] Preferably, target detection in this embodiment of the invention includes: A target detection network (such as the YOLO series network) is used to detect waterbirds in the registered image, and candidate target boxes and their confidence scores are output; when the confidence score of a candidate target is less than a confidence threshold, the target is detected. When this happens, discard the candidate target; The retained target bounding boxes are cropped or feature extracted, and then input into a classification network (such as the ResNet series network) to output species category labels and classification confidence scores, forming a joint "detection-classification" result; Map the center point of the detection box of each candidate target in the image to the panoramic coordinate system or geographic coordinate system, and associate it with the grid number; Because flight path planning involves horizontal and vertical overlap, to avoid duplicate counting, deduplication is performed on targets of the same category within adjacent overlapping areas: the ground distance or panoramic coordinate distance between two targets is calculated, and when the distance is less than a deduplication threshold... If they are identified as the same target, the target with higher confidence level will be retained; The deduplicated target set is summarized by raster number to obtain the machine count results for each raster; further, it is summarized by species to obtain the species count results. Total number of machine counts .
[0040] Preferably, sampling verification includes: Extract a predetermined percentage (e.g., 10%) of the grid cells from the entire grid to form a check set. Optionally, stratified random sampling can be performed after stratifying according to the grid machine counting density to reduce sampling error. right The image or registration result of each raster is manually checked to obtain the raster. Inner species Number of manual reviews And record the corresponding machine count. ; Error coefficients were calculated separately for each species. And correct the total, including: in, For manually reviewed raster sets, For grid Inner species The number of manual reviews For grid Inner species The number of machines counted. For species The total number of machine counts, For the corrected species total.
[0041] When it appears In extreme cases, the species may be included in the "mandatory review" list, increasing the sampling ratio or adopting full manual review.
[0042] Preferably, the target detection network outputs confidence scores for candidate targets and applies a confidence threshold. Perform screening, and when the confidence level of the candidate target is less than The method further includes: discarding the candidate target when it is in time; the method also includes: performing hard example mining on samples with low confidence but high frequency of occurrence, and feeding them back into the training set after manual confirmation, so as to iteratively update the target detection network and / or the classification network.
[0043] Secondly, this invention discloses a system for surveying the concentrated distribution of waterbirds around nuclear power plants based on unmanned aerial vehicles (UAVs), comprising: The drone data acquisition unit is used to fly along the flight path and acquire panoramic and telephoto images, while recording pose information or timestamp information. The route planning unit is used to obtain and rasterize the boundaries of the survey area, based on the lateral overlap. Vertical overlap Generate a zigzag reciprocating flight path and determine the heading step size. ; The registration and embedding unit is used to register the panoramic image and the telephoto image and map the telephoto image to the panoramic coordinate system; The identification and counting unit includes a target detection subunit and a classification subunit. The target detection subunit is used to detect waterbird targets and filter them out according to a confidence threshold, and the classification subunit is used to output the species category. The identification and counting unit is also used to perform deduplication of the same type of target between adjacent overlapping images and to perform machine counting by raster. The sampling correction unit is used to extract a preset proportion of grid cells for verification and calculate the error coefficient to correct the machine count. The output unit is used to output the number and spatial distribution of each species. The units work together to achieve the above method.
[0044] Thirdly, the present invention discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method.
[0045] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the meaning consistent with their meaning in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined.
[0046] For the sake of simplicity, the method embodiments are described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.
[0047] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.
[0048] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for surveying the concentrated distribution of waterbirds around nuclear power plants based on unmanned aerial vehicles (UAVs), characterized in that, include: The survey area was investigated to determine the drone launch points, flight safety constraints, and filming times. Obtain the boundaries of the survey area and rasterize it to divide the survey area into... Each grid, based on horizontal overlap. Vertical overlap Plan a zigzag reciprocating route and determine the heading step size. With the triggering strategy, the coverage of the survey area is not lower than the preset threshold; The drone performs flight photography according to the described route, and acquires image data with panoramic focal length and telephoto focal length respectively based on the same waypoint sequence, and records the pose information or timestamp information of each image; The image data is preprocessed, and the panoramic image and telephoto image are registered, mapping the telephoto image to the panoramic coordinate system. The registered images are used to detect waterbirds using a target detection network and candidate targets are filtered out according to a confidence threshold. The retained targets are then classified into species using a classification network. In adjacent images with overlapping coverage, duplicate targets of the same type are removed to obtain the machine counting results for each grid. A preset proportion of grid cells are extracted from all grid cells for verification, the error coefficient is calculated, the machine count is corrected, and the results of the number of each species and its spatial distribution are output.
2. The method according to claim 1, characterized in that, The rasterization process has a raster horizontal scale of [missing information]. Where W is the effective width of a single image on the ground. and satisfy: in, For flight altitude, The horizontal field of view of the camera. This represents the percentage of invalid edges.
3. The method according to claim 1, characterized in that, The heading step Ground speed from heading Interval between shooting It is certain, and satisfies: and Further satisfy: in, For vertical overlap, This represents the effective coverage length of the image in the heading direction.
4. The method according to claim 1, characterized in that, The coverage rate being not lower than a preset threshold includes: The proportion of the intersection area of the boundary grid with the boundary polygon of the survey area to the grid area is calculated. When the proportion is less than a preset proportion threshold, a boundary reshooting route is generated to compensate for the boundary grid.
5. The method according to claim 1, characterized in that, The aerial photography satisfies low interference constraints, including: When the target bird species is a gull or a shorebird, the flight altitude should not be lower than [missing information]. When the target birds are geese, ducks, or cranes, the flight altitude should not be lower than [missing information]. ; and on the same waypoint sequence, respectively adopt Panoramic focal length and to Two acquisitions are performed at the telephoto focal length to obtain a panoramic image and a telephoto image.
6. The method according to claim 1, characterized in that, The deduplication includes: Map the center point of the detection box of the candidate target in the image to a panoramic coordinate system or a geographic coordinate system. Calculate the pairwise distance between targets of the same category in adjacent overlapping regions. When the distance is less than the deduplication threshold... If the targets are identified as the same target, the target with higher confidence is retained to generate a deduplicated target set.
7. The method according to claim 1, characterized in that, The error coefficient is based on species Calculate separately, and satisfy: in, For manually reviewed raster sets, For grid Inner species The number of manual reviews For grid Inner species The number of machines counted. For species The total number of machine counts, For the corrected species total.
8. The method according to claim 1, characterized in that, The target detection network outputs confidence scores for candidate targets and applies a confidence threshold. Perform screening, and when the confidence level of the candidate target is less than... The method further includes: discarding the candidate target when it is in time; the method also includes: performing hard example mining on samples with low confidence but high frequency of occurrence, and feeding them back into the training set after manual confirmation, so as to iteratively update the target detection network and / or the classification network.
9. A UAV-based system for surveying the concentrated distribution of waterbirds around nuclear power plants, characterized in that, include: The drone data acquisition unit is used to fly along the flight path and acquire panoramic and telephoto images, while recording pose information or timestamp information. The route planning unit is used to obtain and rasterize the boundaries of the survey area, based on the lateral overlap. Vertical overlap Generate a zigzag reciprocating flight path and determine the heading step size. ; The registration and embedding unit is used to register the panoramic image and the telephoto image and map the telephoto image to the panoramic coordinate system; The identification and counting unit includes a target detection subunit and a classification subunit. The target detection subunit is used to detect waterbird targets and filter them out according to a confidence threshold, and the classification subunit is used to output the species category. The identification and counting unit is also used to perform deduplication of the same type of target between adjacent overlapping images and to perform machine counting by raster. The sampling correction unit is used to extract a preset proportion of grid cells for verification and calculate the error coefficient to correct the machine count. The output unit is used to output the number and spatial distribution of each species. The units work together to implement the method according to any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method described in any one of claims 1-8.