Method for extracting behavioral parameters of juvenile Chinese alligators

Abstract

The present invention belongs to the field of Yangtze alligator conservation technology, and discloses a method for extracting behavioral parameters of young Yangtze alligators, comprising the following steps: S1, using a digital camera to continuously record the behavior of young alligators; S2, using a computer to analyze the recorded video of young alligator behavior frame by frame, and identifying the position of the young alligator in each frame image; S3, establishing a two-dimensional matrix containing the position coordinate information of the young alligators as a trajectory pool; S4, starting from the first frame and traversing backwards, when traversing to the i-th frame, extracting the position information of all young alligators in the frame, and matching it with the position coordinates of the current frame of all trajectories in the trajectory pool; S5, if the coordinates of the i-th frame successfully match a certain trajectory in the trajectory pool, then connecting the coordinates with the matching trajectory; otherwise, creating a new trajectory in the trajectory pool with the coordinates as the starting point. The present invention realizes the transition of the behavioral research of young Yangtze alligators from qualitative analysis to large-scale quantitative research, and improves the accuracy of the observation results and the efficiency of data acquisition and processing.

Description

Technical Field

[0001] The invention relates to a Chinese alligator behavior research method, and belongs to the technical field of Chinese alligator conservation. Background Art

[0002] Chinese Alligator ( Alligator sinensis ) is the only species among the 24 existing crocodile species distributed in my country. It is listed as a first-class protected wild animal in my country and is included in the Red List of the World Conservation Union. According to the monitoring data of wild populations, the number of wild Yangtze alligators is still very rare. They are currently only distributed in isolated habitats such as the hilly areas south of the Yangtze River in Anhui and north of the southern Anhui mountain range, the border between Anhui and Jiangsu, and southeastern Anhui. The conservation of Yangtze alligators has been launched since the early 1980s. So far, the scale breeding and breeding techniques have been gradually improved, and the number of artificially bred Yangtze alligators has reached more than 16,000. However, although there are a large number of artificially bred Yangtze alligators, their wild populations are still relatively fragile. The rejuvenation of the Yangtze alligator population is the current focus.

[0003] After the Yangtze alligators hatch, they will be placed in the breeding pond for artificial feeding, and can only be released into the pond after 2-3 years. Taking the Yangtze alligator breeding base of Wuhu Dajiang Industrial Co., Ltd. as an example, the base hatches about 1,000 Yangtze alligators every year. The size of the breeding pond for young alligators in its greenhouse is 80 cm × 40 cm. Generally, more than 20 young alligators are raised in each pond, and the breeding density of young alligators exceeds 60 / m 2 When feeding is unified in the breeding pond, there is a certain degree of competition among the feeding behaviors of the young alligators. Young alligators with poor growth conditions have weak competitiveness and find it difficult to compete for food. In addition, thin young alligators generally adopt a conservative feeding strategy and repeatedly confirm the safety of the environment, which leads to insufficient food intake in the competitive environment of captivity and further deterioration of their growth conditions. Therefore, young alligators with poor growth conditions should be fed in separate basins and given interventional treatment in a timely manner. If they continue to be mixed, they may die within a year. In this case, the death rate of young alligators within a year is usually as high as 20%, which has a great impact on the conservation of Yangtze alligators.

[0004] Since young Yangtze alligators have low blood content, it is difficult to detect their health status by blood sampling. Therefore, it is particularly important to explore and establish a new method to detect sub-healthy alligators early, separate and feed them in pots as soon as possible, and improve the survival rate of young alligators. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned deficiencies in the prior art, and to propose a method for extracting behavioral parameters of young Chinese alligators, so as to facilitate the monitoring of Chinese alligators and the early detection of sub-healthy young alligators, thereby facilitating the subsequent targeted conservation measures to improve the survival rate of young Chinese alligators.

[0006] In order to achieve the above-mentioned object, the present invention provides a method for extracting behavioral parameters of a young Chinese alligator, comprising the following steps: S1, using a digital camera to continuously record the behavior of the young alligator; S2, using a computer to analyze the young alligator behavior video recorded by the digital camera frame by frame, identifying the x-direction and y-direction position coordinates of the young alligator in each frame image and storing them in a text file; S3, reading the young alligator position coordinate data obtained in step S2, generating a three-dimensional data matrix with a first dimension of coordinate number, a second dimension of horizontal and vertical coordinates, and a third dimension of frame number; establishing a two-dimensional matrix containing the horizontal and vertical coordinate information of the young alligator position as a trajectory pool; S4, traversing backward from the first frame, When traversing to the i-th frame, extract the position information of all the young alligators in the frame, and match them with the position coordinates of the current frame of all the tracks in the track pool, calculate the distance between the coordinates, and if the distance between the i-th frame coordinates and the coordinates of a track in the track pool in the current frame is the shortest, and the distance is less than 0.0001, it is determined to be a successful match; S5, if a head coordinate of the i-th frame successfully matches a track in the track pool, connect the head coordinate to the matched track, add 1 to the track length, and update the track pool and track index; if a head coordinate of the i-th frame cannot be successfully matched with any track, create a new track in the track pool with the coordinate as the starting point, update the track pool and establish the track index;

[0007] Perform a secondary connection on the position coordinate information read in S3, and establish a second track pool for storing new track information generated after the secondary connection;

[0008] Traverse all trajectories starting from the first trajectory in the first trajectory pool, calculate the distance between the starting point coordinates of each trajectory and the ending point coordinates of the trajectory in the second trajectory pool, and the frame difference. If the distance between the starting point of the i-th trajectory and the ending point coordinates of a trajectory in the second trajectory pool is less than 0.0001, and the frame difference is less than or equal to 10, then the two trajectories can be successfully matched;

[0009] If the ith trajectory can be successfully matched with a trajectory in the second trajectory pool, the ith trajectory is connected to the end of the trajectory, the coordinates of the missing frames in the middle are filled using interpolation, and the trajectory index of the second trajectory pool is updated according to the connected trajectory. If the ith trajectory cannot be successfully matched with any trajectory in the second trajectory pool, the ith trajectory is stored in the second trajectory pool as an independent new trajectory and an index is established;

[0010] After the trajectory matching is completed, the fragmented trajectories with a length less than 20 are deleted;

[0011] Load the second track pool, and arrange the start and end point frame numbers of all tracks in the second track pool in sequence in the track list; load the original video, and mark the number of the young crocodile to which each track belongs by comparing the original video; extract the coordinates of the four corners of the young crocodile activity area, perform rectangular correction on each coordinate, and then use the length and width of the breeding pool to calibrate the coordinates for actual distance;

[0012] The cumulative distance of the young crocodile's movement is calculated based on the sum of the distances between the young crocodile's position coordinates in two adjacent frames, where

[0013] a. Determination of the coordinates of the baby crocodile's position

[0014] The position coordinates of the young crocodile are calculated from the coordinates of each frame in the complete trajectory of the young crocodile's head and body:

[0015]

[0016]

[0017] In the formula, Px and Py The positions of the young crocodiles x Axis and y Axis coordinates, Headx and Heady Head position x Axis and y Axis coordinates, Bodyx and Bodyy For body position x Axis and y Axis coordinates.

[0018] b. Accumulated distance

[0019] The cumulative distance is the sum of the distances between the baby alligator's position coordinates between every two frames:

[0020]

[0021] Where: S To accumulate distance; N is the total number of frames; Px i and Py i is the position of the young crocodile in the i-th frame x Axis and y Axis coordinates.

[0022] Furthermore, the average number of neighboring companions was calculated based on the number of neighboring companions in a single frame: the number of neighboring companions in a single frame was the number of other young crocodiles whose position coordinates were within a 10 cm diameter range with the target young crocodile position coordinates as the center of the circle in that frame. The average number of neighboring companions was the average number of neighboring companions in all frames.

[0023] Furthermore, after the trajectory matching is completed, the original video is played frame by frame to record the number of frames in which each crocodile actively eats and fights for food.

[0024] Furthermore, a trajectory in the second trajectory pool that actually belongs to different baby crocodiles is disconnected from a certain frame into two trajectories, and then the trajectory pool and the trajectory index are updated, and the baby crocodile number is re-labeled on the disconnected trajectory; the coordinate points of the baby crocodile position are added to the frame with incomplete position coordinates, and then connected with the position coordinates of the same baby crocodile in the previous and next frames to form a complete trajectory.

[0025] Furthermore, in the frame-by-frame analysis, the YOLO-v5 model is used to perform frame-by-frame analysis on the behavior video of the young crocodile recorded by the digital camera through a computer to identify the position of the young crocodile in each frame of the image.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows: the behavior of Yangtze alligator juveniles is recorded by digital equipment, the obtained video is imported into a computer for frame-by-frame analysis, and the position coordinate data of the juvenile crocodile in the breeding pond is obtained by analyzing the position of the juvenile crocodile in the image, so as to facilitate the acquisition of the behavioral data of the juvenile crocodile according to the corresponding position coordinates of the juvenile crocodile in each frame image, and the sub-healthy juvenile crocodiles with abnormal behavior are found by using statistical methods. Traditional research on the behavior of Yangtze alligators juveniles mostly uses manual observation, which is time-consuming and labor-intensive, and difficult to conduct precise quantitative analysis in batches. The present invention establishes a model to perform statistical analysis on the collected digital images of the behavior of juvenile crocodiles, realizing the transition of the behavior research of Yangtze alligators juveniles from qualitative analysis to large-scale quantitative research, which not only eliminates the error of manual observation, greatly improves the accuracy of the observation results, but also reduces the intensity of observation research work. At the same time, the digital image can quickly extract the behavioral parameter data of the juvenile crocodiles through computer batch processing, making the behavior research of Yangtze alligators juveniles digital and intelligent. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A schematic diagram of a step of acquiring trajectory data of a young crocodile in an embodiment of the present invention;

[0028] Figure 2 Schematic diagram of another step of acquiring trajectory data of a young crocodile in an embodiment of the present invention;

[0029] Figure 3 It is a two-dimensional schematic diagram of the displacement data extraction result of a young crocodile in an embodiment of the present invention;

[0030] Figure 4 It is a schematic diagram of the cumulative moving distance and food snatching behavior extraction results of a young crocodile in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments, and should not be understood as limiting the scope of protection of the present invention.

[0032] See also Figure 1 The method for extracting behavioral parameters of a young Chinese alligator in an embodiment of the present invention includes the following steps: frame-by-frame recognition of the head and body positions of the Chinese alligator - preliminary connection of trajectories - secondary connection of trajectories - verification of trajectories using tools - marking of the young alligator's eating behavior using tools.

[0033] (1) Use the YOLO-v5 model to identify the coordinates of the head and body parts of the young alligator in each frame of the video

[0034] YOLO-v5 is an object detection model designed based on computer vision patterns. It is an upgraded version of the YOLO series, which is known for its real-time object detection capabilities. The established YOLO-v5 model is used to identify the motion video of the Chinese alligator frame by frame, and the position coordinates of the head and body parts of the young alligator in each frame of the video (the coordinates of the center point of the marking box) will be stored in a text file respectively.

[0035] (2) Use the trajectory preliminary connection method to preliminarily connect the head and body position coordinates of each frame into a trajectory

[0036] The preliminary track connection method is divided into the preliminary track connection of the head and the body. The method flow is as follows: Figure 1 As shown in the figure, since the two connection methods of the head and body tracks are basically the same, the following will only introduce the method of the head track.

[0037] 2.1 Data reading and trajectory pool establishment

[0038] First, the coordinates of the head and body positions of the young alligator in all frames output by the YOLO-v5 model are read from the file and organized into a three-dimensional data matrix with the first dimension being the number of coordinates, the second dimension being the horizontal and vertical coordinates, and the third dimension being the number of frames. Then a two-dimensional matrix is ​​established as a track pool, with the horizontal and vertical coordinates of a track placed in every two rows, and an index file is established to store the starting point, end point, length, and other information of each track.

[0039] 2.2 Traversal matching of each frame data

[0040] Start from frame 1 and traverse backwards. i When a frame is generated, all head positions in the frame are extracted and matched with the head position coordinates of all tracks in the track pool in the current frame, and the distance between the two coordinates is calculated. iIf the distance between the head coordinates in the frame and the coordinates of a track in the track pool in the current frame is the shortest, and the distance is less than 0.0001 (normalized coordinate distance), then it is considered a successful match.

[0041] 2.3 Track Pool Update

[0042] After the matching is completed, if i If a head coordinate in a frame matches a track in the track pool, the head coordinate is connected to the matching track, the track length is increased by 1, and the track pool and track index are updated. i If the head coordinates of a frame cannot be matched with any track, a new track with the coordinates as the starting point is created in the track pool, and the track pool is updated and the track index is established. i =1, there is no track in the track pool, and all head coordinates will be used as the starting point of the track.

[0043] After traversing all frames of the video using the above method, the extracted trajectory coordinate data is stored in the trajectory pool, and each trajectory data can be found using the trajectory index.

[0044] (3) Use the trajectory secondary connection method to further combine the trajectories

[0045] Since the YOLO-v5 model may have a certain error rate in recognizing the head and body of the young crocodile, the head or body trajectory in some frames is not recognized, which will cause the trajectory obtained by the initial connection to be disconnected. Therefore, a trajectory secondary connection method is developed to further connect the trajectories. The method flow is as follows: Figure 2 As shown in the figure, since the secondary connection methods of the head and body tracks are basically the same, the following only introduces the secondary connection method of the head track.

[0046] 3.1 Establishment of the Second Trajectory Pool

[0047] First, a second track pool is established to store the tracks after the second connection. The index format of the second track pool is consistent with the track pool obtained by the initial connection method.

[0048] 3.2 Traversal and merging of trajectory data

[0049] Traverse all tracks starting from the first track in the original track pool, and calculate the distance between the starting point coordinates of each track and the ending point coordinates of the track in the second track pool (normalized coordinate distance) as well as the frame difference. i If the coordinate distance between the starting point of a trajectory and the ending point of a trajectory in the second trajectory pool is less than 0.0001, and the frame difference is less than or equal to 10, then the two trajectories can be successfully matched.

[0050] 3.3 Update of the Second Track Pool

[0051] After the matching is completed, if i If the first track can be successfully matched with a track in the second track pool, i The track is connected to the end of the track, the coordinates of the missing frames in the middle are filled using interpolation, and the track index of the second track pool is updated according to the connected track. i The trajectory cannot be successfully matched with any trajectory in the second trajectory pool, then i The trajectory is stored as an independent new trajectory in the second trajectory pool and an index is created.

[0052] After the above secondary connection, the obtained trajectory tends to be complete. Then the fragmented trajectories with a length less than 20 are deleted, which are mostly caused by the misjudgment of the model.

[0053] (4) Use software to manually correct the trajectory

[0054] If there are 1 or 2 young alligators in the video, the trajectory of the young alligators can be well obtained through the initial and secondary connection of the trajectory. However, when the number of young alligators increases, it is very easy for multiple young alligators to cross each other, and the bodies and heads of different young alligators overlap, which will cause errors in the trajectory automatically connected by the algorithm, such as the front and back parts of a trajectory belonging to different young alligators, or the trajectory is broken due to the long-term mutual occlusion of young alligators. For this situation, a trajectory inspection and feeding behavior marking tool was developed using the GUI tool of Matlab 2014 software to correct the incorrectly connected trajectories and mark the time when each young alligator's feeding behavior occurred.

[0055] 4.1 Loading the head or body trajectory after secondary connection

[0056] Load the head or body track pool after secondary connection, and arrange the start and end frame numbers of all tracks in the pool in sequence in the track list. Click the frame number in the list to display the coordinates of the start and end points of the track, and display the complete path of the track in the track display window.

[0057] 4.2 Trajectory Classification

[0058] The track after the second connection is an unlabeled track and cannot be matched to the individual crocodile. Each track needs to be marked with the crocodile number. After loading the head or body track using the software, load the original video and manually compare and identify the crocodile number to which each track belongs. After marking the number, the tracks of different individuals will be displayed in different colors.

[0059] 4.3 Trajectory Segmentation

[0060] Due to the mutual crossing of young alligators, there may be connection errors at the intersection of the tracks, and one track may belong to different young alligators. In this case, you need to use the track segmentation function of the software tool to split a track from a certain frame into two tracks, and then update the track pool and track index, and then you can mark the young alligator number on the disconnected track through the software.

[0061] 4.4 Adding new track points

[0062] Due to model recognition errors or deletion of too short trajectories, some crocodile trajectories will be interrupted. In this case, you need to use the software's function of adding trajectory points to manually add coordinate points of the crocodile's head and body positions for frames with incomplete position coordinates. Click the trajectory point to select the crocodile number, and click the crocodile's head or body response position in the video display frame to obtain the coordinate information and automatically add it.

[0063] 4.5 Complete trajectory generation

[0064] After the juvenile crocodile trajectory is segmented, reconnected and dotted, the complete trajectory generation tool can be used to generate a complete head or body trajectory, the missing values ​​in each juvenile crocodile trajectory are filled using the interpolation method, and the head and body trajectories of each juvenile crocodile are denoised using the median filter algorithm with a window of 20. Then the coordinates of the four corners of the juvenile crocodile activity area are extracted, each coordinate is rectangularly corrected, and the actual distance of the coordinates is calibrated using the length and width of the breeding pond.

[0065] (5) Using tools to mark eating behavior

[0066] After the complete head and body tracks were generated, the eating behavior marking tool was used to load the complete head track, body track and original video to further verify the correctness of the track connection. Then the original video was played frame by frame to record the number of frames in which each crocodile showed active eating and food fighting behaviors.

[0067] (6) Calculation of various behavioral characteristic parameters

[0068] a. Determination of the coordinates of the baby crocodile's position

[0069] The position coordinates of the young crocodile are calculated from the coordinates of each frame in the complete trajectory of the young crocodile's head and body:

[0070]

[0071]

[0072] In the formula, Px and Py The positions of the young crocodiles x Axis and y Axis coordinates, Headx and HeadyHead position x Axis and y Axis coordinates, Bodyx and Bodyy For body position x Axis and y Axis coordinates.

[0073] b. Accumulated distance

[0074] The cumulative distance is the sum of the distances between the baby alligator's position coordinates between every two frames:

[0075]

[0076] Where: S To accumulate distance; N is the total number of frames; Px i and Py i is the position of the young crocodile in the i-th frame x Axis and y Axis coordinates.

[0077] c. Eating and snatching times

[0078] The number of times the animal eats and grabs food is obtained by adding up the manually calibrated data.

[0079] d. Average number of neighboring peers

[0080] The number of neighboring companions in a single frame is the number of other young alligators whose position coordinates are within a 10 cm diameter range with the target young alligator as the center. The average number of neighboring companions is the average number of neighboring companions in all frames.

[0081] In order to test the effectiveness of the developed model for extracting behavioral parameters of young crocodiles, the behavioral extraction method was verified using a feeding video of newborn young crocodiles on September 24, 2021.

[0082] First, the YOLO-v5 convolutional neural network model was used to identify the position of the head and body of the young alligator in the video frame by frame; then, the trajectory preliminary connection algorithm and trajectory secondary connection algorithm were used to connect the head and body trajectories respectively; the young alligator trajectory inspection and eating behavior marking tool was used to verify the trajectory and mark the young alligator's active eating and food snatching behavior; finally, the trajectory, displacement, average number of neighboring companions, and the number and time of eating behavior of each young alligator were extracted. The data extraction results are shown in Table 1. Figure 3 and Figure 4 .

[0083] The extraction effect of the baby crocodile's moving trajectory is as follows Figure 3As shown in the figure, it can be seen that this method can extract the movement trajectory of the young crocodile very accurately, and can distinguish it from the trajectory of other young crocodiles through color marking. Figure 4 As shown, it can be seen that this method can extract the cumulative moving distance of each young crocodile, and use time as the axis to mark the time when each young crocodile actively eats and grabs food.

[0084] Table 1 Results of extracting behavioral characteristic parameters of young crocodiles

[0085]

[0086] The extraction results of this method for each young crocodile, including the total displacement distance (unit: cm), number of times eating, number of times snatching food, and average number of neighboring companions, are shown in Table 1.

[0087] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technical solution obtained by any technician familiar with the technical field within the technical scope disclosed by the present invention and by equivalent replacement or modification based on the concept and technical principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for extracting behavioral parameters of young Chinese alligators, characterized in that: The method comprises the following steps: S1, continuously recording the behavior of the young crocodile by using a digital camera; S2, analyzing the video of the behavior of the young crocodile recorded by the digital camera frame by frame by a computer, identifying the x-direction and y-direction position coordinates of the young crocodile in each frame image and storing them in a text file; S3, reading the young crocodile position coordinate data obtained in step S2, generating a three-dimensional data matrix with the first dimension being the number of coordinates, the second dimension being the horizontal and vertical coordinates, and the third dimension being the number of frames; and establishing a two-dimensional matrix containing the horizontal and vertical coordinate information of the young crocodile position as a trajectory pool; S4, traverse backward from the 1st frame, when traversing to the i-th frame, extract the position information of all the young alligators in the frame, and match them with the position coordinates of the current frame of all the tracks in the track pool, calculate the distance between the coordinates, and if the distance between the coordinates of the i-th frame and the coordinates of the current frame of a track in the track pool is the shortest, it is determined to be a successful match; S5, if a head coordinate of the i-th frame successfully matches a track in the track pool, connect the head coordinate with the matched track, add 1 to the length of the track, and update the track pool and track index; If a head coordinate in the i-th frame cannot be successfully matched with any trajectory, a new trajectory with the coordinate as the starting point is created in the trajectory pool, and the trajectory pool is updated and the trajectory index is established; Perform a secondary connection on the position coordinate information read in S3, and establish a second track pool for storing new track information generated after the secondary connection; Traverse all trajectories starting from the first trajectory in the first trajectory pool, calculate the distance between the starting point coordinates of each trajectory and the ending point coordinates of the trajectory in the second trajectory pool, and the frame difference. If the distance between the starting point of the i-th trajectory and the ending point coordinates of a trajectory in the second trajectory pool is less than 0.0001, and the frame difference is less than or equal to 10, then the two trajectories can be successfully matched; If the ith trajectory can be successfully matched with a trajectory in the second trajectory pool, the ith trajectory is connected to the end of the trajectory, the coordinates of the missing frames in the middle are filled using interpolation, and the trajectory index of the second trajectory pool is updated according to the connected trajectory. If the ith trajectory cannot be successfully matched with any trajectory in the second trajectory pool, the ith trajectory is stored in the second trajectory pool as an independent new trajectory and an index is established; After the trajectory matching is completed, the fragmented trajectories with a length less than 20 are deleted; Load the second track pool, and arrange the start and end point frame numbers of all tracks in the second track pool in sequence in the track list; load the original video, and mark the number of the young crocodile to which each track belongs by comparing the original video; extract the coordinates of the four corners of the young crocodile activity area, perform rectangular correction on each coordinate, and then use the length and width of the breeding pool to calibrate the coordinates for actual distance; The cumulative distance of the young crocodile is calculated based on the sum of the distances between the young crocodile's position coordinates in two adjacent frames, where: a. Determination of the coordinates of the baby crocodile's position The position coordinates of the young crocodile are calculated from the coordinates of each frame in the complete trajectory of the young crocodile's head and body: ； ； In the formula, Px and Py The positions of the young crocodiles x Axis and y Axis coordinates, Headx and Heady Head position x Axis and y Axis coordinates, Bodyx and Bodyy For body position x Axis and y Axis coordinates; b. Accumulated distance The cumulative distance is the sum of the distances between the baby alligator's position coordinates between every two frames: ； Where: S To accumulate distance; N is the total number of frames; Px i and Py i is the position of the young crocodile in the i-th frame x Axis and y Axis coordinates.

2. The method for extracting behavioral parameters of young Chinese alligators according to claim 1, characterized in that: The average number of neighboring companions was calculated based on the number of neighboring companions in a single frame: the number of neighboring companions in a single frame was the number of other young crocodiles whose position coordinates were within a 10 cm diameter range with the target young crocodile position coordinates as the center of the circle in that frame; the average number of neighboring companions was the average number of neighboring companions in all frames.

3. The method for extracting behavioral parameters of young Chinese alligators according to claim 1, characterized in that: After the trajectory matching is completed, the original video is played frame by frame to record the number of frames in which each crocodile actively eats and fights for food.

4. The method for extracting behavioral parameters of young Chinese alligators according to claim 1, characterized in that: A track in the second track pool that actually belongs to different baby crocodiles is split into two tracks from a certain frame, and then the track pool and track index are updated, and the baby crocodile number is re-labeled on the split track; the coordinate points of the baby crocodile position are added to the frames with incomplete position coordinates, and then connected with the position coordinates of the same baby crocodile in the previous and next frames to form a complete track.

5. The method for extracting behavioral parameters of young Chinese alligators according to any one of claims 1 to 4, characterized in that: In the frame-by-frame analysis, the YOLO-v5 model is used to perform frame-by-frame analysis on the behavior video of the young crocodile recorded by the digital camera through a computer to identify the position of the young crocodile in each frame of the image.

Images