Northeast tiger individual identification method fusing spatio-temporal information

Abstract

A northeast tiger individual identification method fusing spatio-temporal information belongs to the technical field of artificial intelligence, and comprises the following steps: collecting and marking a video of a northeast tiger individual shot by a wild protection camera, body side pattern area information, key point information, body orientation, field protection camera number and time information of the northeast tiger individual are obtained; acquiring an image sequence of a body side area of a northeast tiger individual based on a northeast tiger target detector, and estimating the body orientation of the northeast tiger individual based on a posture recognition algorithm; predicting the individual confidence coefficient of the northeast tiger based on the body side pattern features; and determining Northeast tiger individual information by combining the Northeast tiger spatio-temporal information The method solves the problems that in an existing northeast tiger individual identification method, motion blurring and tree shielding exist in a single-frame image of a northeast tiger individual shot by a field protection camera in an actual operation scene, and domain features, positions and time information of the northeast tiger are not fully utilized, and has the advantages of being high in detection precision, high in robustness and the like.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to a method for identifying individual Amur tigers by integrating spatio-temporal information. Background Art

[0002] In recent years, due to factors such as habitat loss, illegal hunting, and human activity interference, the number of Amur tigers has decreased sharply, and it has been listed as a globally critically endangered species. In order to reverse the decline trend of the Amur tiger population, the government and environmental protection organizations have launched a number of population restoration plans. Through individual monitoring, the survival status of Amur tigers can be understood in a timely manner, providing data support for formulating effective protection measures.

[0003] Currently, the method for identifying individual Amur tigers based on computer vision uses the unique identification feature of the body side patterns of different Amur tigers. First, the body side pattern area is extracted through a target detection network, and then the individual identification result of the Amur tiger is obtained through the prediction of a neural network.

[0004] Chinese Patent CN114220120A, "A Method for Identifying Individual Amur Tigers", first extracts the pattern images of the trunk and hind legs to construct a reference database for individual Amur tigers, and trains a classifier based on a dense convolutional neural network in the reference database; then uses the YOLO target detection method based on CNN to detect Amur tiger targets in real time, and uses an animal pose estimation method to extract the real-time pose of the Amur tiger target, and confirms the body orientation of the Amur tiger according to the pose information; finally, automatically extracts the pattern image sequence of the trunk and hind legs of the Amur tiger based on the real-time detection result, and obtains the final individual classification result of the Amur tiger through classifier prediction and fusion. However, wildlife protection cameras are usually deployed in the forest. On the one hand, there is a phenomenon that the body side of the captured Amur tiger is blocked by trees. As the body side pattern is the only identification feature of the Amur tiger individual, only performing neural network prediction analysis on a single-frame image will affect the accuracy of model recognition; on the other hand, Amur tigers have a strong territorial awareness, specifically manifested as: when the wildlife protection camera is in the territory of an individual Amur tiger, there is a very small probability of capturing other Amur tiger individuals. Therefore, using the spatio-temporal information of the appearance of Amur tiger individuals helps to improve the effect of individual identification.

[0005] In summary, the existing methods for identifying individual Amur tigers have the following problems:

[0006] 1) In the actual operation scenario of wildlife protection cameras, the captured Amur tiger individuals have problems of motion blur and tree occlusion in a single-frame image.

[0007] 2) The territorial characteristics, location, and time information of Amur tigers are not fully utilized. Summary of the Invention

[0008] To solve the problems existing in the existing individual recognition methods of Amur tigers, such as motion blur, tree occlusion of Amur tigers captured by wildlife protection cameras in actual operation scenarios, and the failure to fully utilize the territorial characteristics, location, and time information of Amur tigers, the present invention provides an individual recognition method for Amur tigers that integrates spatio-temporal information.

[0009] The technical solutions adopted by the present invention to solve the technical problems are specifically as follows:

[0010] An individual recognition method for Amur tigers that integrates spatio-temporal information provided by the present invention specifically includes the following steps:

[0011] Step 1: Collect and annotate the videos of Amur tigers captured by wildlife protection cameras to obtain the flank pattern area information of Amur tigers, the key point information of Amur tigers, the body orientation of Amur tigers, the wildlife protection camera number, and time information.

[0012] Step 2: Based on the Amur tiger target detector, obtain the image sequence of the flank area of the Amur tiger individual, and estimate the body orientation of the Amur tiger individual based on the pose recognition algorithm.

[0013] Step 3: Predict the confidence of the Amur tiger individual based on the flank pattern features.

[0014] Step 4: Combine the spatio-temporal information of the Amur tiger to judge the individual information of the Amur tiger.

[0015] Further, in Step 2, first construct an Amur tiger target detection dataset according to the flank pattern area information of the Amur tiger individual obtained in Step 1, and train an Amur tiger target detector on the Amur tiger target detection dataset; then use the Amur tiger target detector to predict the coordinate information of the Amur tiger individual in the video; finally, intercept the image sequence of the flank area of the Amur tiger individual from the video based on the coordinate information.

[0016] Further, in Step 2, first construct an Amur tiger key point detection dataset according to the key point information of the Amur tiger individual obtained in Step 1, and train an Amur tiger key point detector on the Amur tiger key point detection dataset; then use the Amur tiger key point detector to predict the middle frame of the Amur tiger individual video to obtain the key point information of the Amur tiger individual; finally, judge the body orientation of the Amur tiger individual according to the key point information of the Amur tiger individual using the pose recognition algorithm.

[0017] Further, the specific implementation process of the pose recognition algorithm is as follows:

[0018] Calculate the mean value L of the abscissas of the head key points 1, 2, and 3 of the Amur tiger in the image head , and the mean value L of the abscissas of the hind limb key points 9 to 15 in the image tail , when L head < Ltail When L head >L tail , its body is judged to be facing the right side.

[0019] Furthermore, in step three, T frame images are first sampled from the image sequence of the lateral area of ​​the Siberian tiger individual obtained in step two, and then resized to the training size after letterbox operation to establish a Siberian tiger individual identification dataset; then a multi-frame Siberian tiger individual identification model based on convolutional neural network and graph neural network is constructed and trained to obtain a Siberian tiger individual identification model.

[0020] Furthermore, the Siberian tiger individual recognition model includes: an intra-frame feature extraction module based on a convolutional neural network, a multi-frame feature aggregation module based on a graph convolutional neural network, and a classification module; the intra-frame feature extraction module extracts features from each frame in the input Siberian tiger individual body side area image sequence to obtain a corresponding feature map sequence; the obtained feature map is divided into blocks and a data map structure is constructed based on similarity; the multi-frame feature aggregation module interactively learns these feature blocks to obtain fused features; and the confidence information of the Siberian tiger individual is obtained through the classification module.

[0021] Furthermore, the specific implementation process of step 4 is as follows:

[0022] S4.1: Use the wildlife camera numbers and time information obtained in step 1 to build a temporal and spatial information database of Siberian tiger individuals;

[0023] S4.2: Define the set of wildlife cameras covering Siberian tiger individuals from nt-1 to n-1 weeks is the territory of the nth week, and C represents the complete set of all cameras, that is, Indicates the camera number from 1 to N c The collection of N c represents the number of wildlife cameras; the probability P(n) of the migration of individual Siberian tiger territories in the nth week is counted;

[0024] S4.3: Let the wildlife camera numbered j capture a Siberian tiger as event A j , the Siberian tiger with individual number i was photographed as event B i , the probability of a wild animal camera numbered j capturing an individual Siberian tiger is P(A j ) and the probability P(A) that a Siberian tiger with the number i is photographed by a wildlife camera with the number j j |B i );

[0025] S4.4: Based on the wildlife protection camera number and time information when the Amur tiger individual is photographed, predict the probability of the Amur tiger individual by fusing the spatio-temporal information of the Amur tiger individual;

[0026] S4.5: Judge the Amur tiger individual information according to the probability information of the Amur tiger individual calculated after fusing the spatio-temporal information;

[0027] S4.6: Update the judged Amur tiger individual information to the spatio-temporal information database of the Amur tiger individual, and update the probability P(n) of the migration of the Amur tiger individual's territory every week.

[0028] Furthermore, the calculation formula for the probability P(n) of the migration of the Amur tiger individual's territory in the nth week is:

[0029]

[0030] where N t represents the number of Amur tiger individuals; s represents the event that an Amur tiger individual is photographed within the territory; represents the number of times the Amur tiger with individual number i is photographed within the territory in the nth week; C j (n) represents the number of Amur tigers photographed by the wildlife protection camera numbered j in the nth week.

[0031] Furthermore, the calculation formulas for the probability P(A j ) that the wildlife protection camera numbered j captures an Amur tiger individual event and the probability P(A j |B i ) that the Amur tiger with individual number i is captured by the wildlife protection camera numbered j are respectively:

[0032]

[0033] where represents the number of times the wildlife protection camera numbered j captures the Amur tiger with individual number i; N t represents the number of Amur tiger individuals.

[0034] Furthermore, the calculation formula for the probability of the Amur tiger individual is:

[0035]

[0036] where P(i) represents the probability that the Amur tiger with individual number i is captured by the wildlife protection camera numbered j in the nth week; P(B i ) represents the probability that the Amur tiger with individual number i appears, that is, the confidence information of the Amur tiger individual predicted by the Amur tiger individual recognition model in step three; N t represents the number of Amur tiger individuals;

[0037] The calculation formula for the individual information of the Amur tiger is as follows:

[0038] Tiger id =argmax(P(i)), i ∈ (1, N t )

[0039] where Tiger id represents the result of the individual number of the Amur tiger finally predicted; P(i) represents the probability that the Amur tiger individual belongs to the number i; argmax(P(i)) represents the individual number corresponding to the predicted maximum probability value. The beneficial effects of the present invention are:

[0040] Compared with the existing Amur tiger individual recognition methods, the Amur tiger individual recognition method integrating spatio-temporal information provided by the present invention brings the following improvements to the Amur tiger individual recognition method:

[0041] 1. Strong robustness: The method provided by the present invention samples and extracts frames from the obtained Amur tiger individual video into an image sequence, and predicts the confidence of the Amur tiger individual by aggregating multi-frame body side feature information through a graph neural network. Compared with the neural network learning and predicting a single-frame image, this method can effectively alleviate the influence of motion blur and tree occlusion in a single-frame image on recognition.

[0042] 2. High detection accuracy: The present invention constructs a spatio-temporal information database of Amur tiger individuals based on the characteristics of the territorial awareness of Amur tigers. On the one hand, it provides effective data support for analyzing the characteristics of Amur tiger territories and territorial migrations, and on the other hand, it effectively improves the accuracy of Amur tiger individual recognition.

[0043] In summary, the Amur tiger individual recognition method integrating spatio-temporal information provided by the present invention solves the problems existing in the existing Amur tiger individual recognition methods, such as motion blur, tree occlusion in a single-frame image of the Amur tiger individuals captured by wildlife protection cameras in the actual operation scenario, and the lack of full utilization of the territorial characteristics, positions, and time information of Amur tigers. It has the advantages of high detection accuracy, strong robustness, etc., and provides favorable data support for timely understanding the survival status of Amur tigers and formulating effective protection measures in a timely manner. Brief Description of the Drawings

[0044] Figure 1 is a flowchart of the Amur tiger individual recognition method integrating spatio-temporal information provided by the present invention.

[0045] Figure 2 is a structural diagram of the Amur tiger individual recognition model constructed in step S3.3 of the present invention.

[0046] Figure 3 are 15 joint points marked according to the joint positions in the middle frame of the Amur tiger individual video. Detailed implementation mode

[0047] The present invention provides a method for identifying individual Amur tigers by fusing spatio-temporal information. First, the lateral area of the Amur tiger in the video is extracted and sampled into an image sequence. Then, a frame feature extraction module based on a convolutional neural network is used to extract the feature information of the lateral side of the Amur tiger in each frame of the image sequence, and a multi-frame feature aggregation module based on a graph neural network is used to aggregate the multi-frame lateral feature information of the Amur tiger individual, and then the confidence of the Amur tiger individual is predicted. Finally, a spatio-temporal information database of Amur tiger individuals is constructed, and the individual information of the Amur tiger is jointly judged based on the confidence and spatio-temporal information of the Amur tiger individual.

[0048] Next, in combination with the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0049] A method for identifying individual Amur tigers by fusing spatio-temporal information provided in this embodiment is specifically implemented as follows:

[0050] Step 1: Collect and label the videos of individual Amur tigers captured by wildlife protection cameras. The specific operation steps are as follows:

[0051] S1.1: Collect the videos of individual Amur tigers captured by wildlife protection cameras.

[0052] S1.2: Label the lateral pattern area of the Amur tiger individual in the video of the Amur tiger individual.

[0053] S1.3: Label the key point information and body orientation of the Amur tiger individual in the video of the Amur tiger individual, and set the individual number of the Amur tiger.

[0054] Label the key point information of the Amur tiger individual in the video of the Amur tiger individual. Specifically, as Figure 3 shown, select the middle frame of the video and label 15 key points according to the joint positions.

[0055] S1.4: Count the wildlife protection camera numbers and time information corresponding to the videos of individual Amur tigers.

[0056] Step 2: Obtain an image sequence of the lateral area of the Amur tiger individual based on an Amur tiger target detector, and estimate the body orientation of the Amur tiger individual based on a pose recognition algorithm. The specific operation steps are as follows:

[0057] S2.1: First, construct a Siberian tiger target detection dataset based on the flank pattern area information of the Siberian tiger individuals marked in step S1.2, and train a Siberian tiger target detector on the Siberian tiger target detection dataset; then, predict the coordinate information of the Siberian tiger individuals in the video based on the Siberian tiger target detector; finally, intercept the image sequence of the flank area of the Siberian tiger individuals from the video based on the coordinate information.

[0058] S2.2: First, construct a Siberian tiger key point detection dataset based on the key point information of the Siberian tiger individuals marked in step S1.3, and train a Siberian tiger key point detector on the Siberian tiger key point detection dataset; then, predict the middle frame of the video of the Siberian tiger individuals based on the Siberian tiger key point detector, and predict the key point information (15 joints) of the Siberian tiger individuals; finally, judge the body orientation of the Siberian tiger individuals according to the key point information of the Siberian tiger individuals using the pose recognition algorithm.

[0059] Among them, the pose recognition algorithm is only used to recognize the left - right orientation of the Siberian tiger individuals. According to the relative position relationship of the 15 key points predicted by the key point detector, the body orientation is judged.

[0060] The specific implementation process of the pose recognition algorithm is as follows:

[0061] As Figure 3 shown, calculate the mean value L of the abscissas of the head key points 1, 2, and 3 of the Siberian tiger in the image head , and the mean value L of the abscissas of the hind limb key points 9 to 15 in the image tail . When L head < L tail , judge that its body orientation is to the left; when L head > L tail , judge that its body orientation is to the right.

[0062] Step three: Predict the confidence of the Siberian tiger individuals based on the flank pattern features; the specific operation steps are as follows:

[0063] S3.1: Construct a Siberian tiger individual recognition dataset based on the image sequence of the flank area of the Siberian tiger individuals extracted in step S2.1; specifically, sample and extract T frames of images from the image sequence of the flank area of the Siberian tiger individuals, resize them to the training size after the letterbox operation, and establish a Siberian tiger individual recognition dataset.

[0064] S3.2: Construct a multi - frame Siberian tiger individual recognition model based on the convolutional neural network and the graph neural network.

[0065] S3.3: Train the multi - frame Siberian tiger individual recognition model on the constructed Siberian tiger individual recognition dataset to obtain the final Siberian tiger individual recognition model.

[0066] Specifically, Figure 2 As shown in the figure, the constructed Siberian tiger individual recognition model mainly includes an intra-frame feature extraction module based on convolutional neural network, a multi-frame feature aggregation module based on graph convolutional neural network, and a classification module. First, the intra-frame feature extraction module extracts features from each frame in the input image sequence of the lateral area of ​​the Siberian tiger individual to obtain the corresponding feature map sequence; secondly, the obtained feature map is divided into blocks and the data graph structure is constructed based on the similarity; then the multi-frame feature aggregation module interactively learns these feature blocks to obtain fusion features; finally, the confidence information of the Siberian tiger individual is obtained through the classification module.

[0067] Step 4: Combine the temporal and spatial information of the Siberian tiger to determine the individual information of the Siberian tiger; the specific operation steps are as follows:

[0068] S4.1: Building a temporal and spatial information database of Siberian tiger individuals based on the labeling information (wildlife camera number and time information) in step S1.4;

[0069] S4.2: Define the set of wildlife cameras covering Siberian tiger individuals from nt-1 to n-1 weeks is the territory of the nth week, and C represents the complete set of all cameras, that is, Indicates the camera number from 1 to N c The probability P(n) of the migration of individual Siberian tiger territories in the nth week is calculated. The specific calculation formula is as follows:

[0070]

[0071] Where n is the week number; i is the individual number of the Siberian tiger; j is the wildlife camera number; N t Represents the number of Siberian tiger individuals; N c represents the number of wildlife cameras; s represents the number of Siberian tigers photographed in the territory; represents the number of times the Siberian tiger with individual number i was photographed in the territory in week n; C j (n) represents the number of Siberian tigers photographed by wildlife camera numbered j in week n.

[0072] S4.3: Let the wildlife camera numbered j capture a Siberian tiger as event A j , the Siberian tiger with individual number i was photographed as event B i , the probability of a wild animal camera numbered j capturing an individual Siberian tiger is P(A j ) and the probability P(A) that a Siberian tiger with the number i is photographed by a wildlife camera with the number j j |B i ), the specific calculation formula is as follows:

[0073]

[0074] Among them, i represents the individual number of the Amur tiger; j represents the number of the wildlife protection camera; represents the number of times the Amur tiger with individual number i is photographed by the wildlife protection camera numbered j; N t represents the number of Amur tiger individuals; N c represents the number of wildlife protection cameras.

[0075] S4.4: Based on the number of the wildlife protection camera and the time information when the Amur tiger individual is photographed, predict the probability of the belonging Amur tiger individual by fusing the spatio-temporal information of the Amur tiger individual. The calculation formula is as follows:

[0076]

[0077] Among them, i represents the individual number of the Amur tiger; j represents the number of the wildlife protection camera; P(i) represents the probability that the Amur tiger individual numbered i is photographed by the wildlife protection camera numbered j in the nth week; P(n) represents the probability of the migration of the territory of the Amur tiger individual in the nth week; P(B i ) represents the probability of the appearance of the Amur tiger with individual number i, that is, the confidence information of the Amur tiger individual predicted by the Amur tiger individual recognition model in step three; P(A j ) represents the probability of the event that the wildlife protection camera numbered j photographs the Amur tiger individual; P(A j |B i ) represents the probability of the event that the Amur tiger with individual number i is photographed by the wildlife protection camera numbered j; N t represents the number of Amur tiger individuals.

[0078] S4.5: Judge the Amur tiger individual information according to the probability information of the Amur tiger individual calculated after fusing the spatio-temporal information. The specific calculation formula is as follows:

[0079] Tiger id = argmax(P(i)), i ∈ (1, N t )

[0080] Among them, i represents the individual number of the Amur tiger; N t represents the number of Amur tiger individuals; Tiger id represents the final predicted result of the individual number of the Amur tiger; P(i) represents the probability that the Amur tiger individual belongs to number i; argmax(P(i)) represents the individual number corresponding to the predicted maximum probability value.

[0081] S4.6: Update the judged Amur tiger individual information to the Amur tiger individual spatio-temporal information database, and update the probability P(n) of the migration of the territory of the Amur tiger individual every week.

[0082] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for identifying Siberian tiger individuals by integrating spatiotemporal information, characterized in that: The following steps are involved: Step 1: Collect and annotate videos of individual Siberian tigers captured by wildlife cameras, obtain information about the side pattern area of ​​the individual Siberian tigers, key point information of the individual Siberian tigers, body orientation of the individual Siberian tigers, wildlife camera numbers, and time information; Step 2: Obtain an image sequence of the side area of ​​the Siberian tiger individual based on the Siberian tiger target detector, and estimate the body orientation of the Siberian tiger individual based on the posture recognition algorithm; Step 3: Predict the individual confidence of Siberian tiger based on the characteristics of body side pattern; Step 4: Combine the temporal and spatial information of the Siberian tiger to determine the individual information of the Siberian tiger.

2. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 1, characterized in that: In step 2, a Siberian tiger target detection dataset is first constructed based on the side body pattern area information of the Siberian tiger individual obtained in step 1, and a Siberian tiger target detector is obtained by training on the Siberian tiger target detection dataset; the Siberian tiger target detector is then used to predict the coordinate information of the Siberian tiger individual in the video; finally, an image sequence of the side body area of ​​the Siberian tiger individual is captured from the video based on the coordinate information.

3. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 1, characterized in that: In step 2, a Siberian tiger key point detection dataset is first constructed based on the key point information of the Siberian tiger individuals obtained in step 1, and a Siberian tiger key point detector is obtained by training on the Siberian tiger key point detection dataset; the Siberian tiger key point detector is then used to predict the intermediate frames of the Siberian tiger individual video to obtain the key point information of the Siberian tiger individual; finally, the body orientation of the Siberian tiger individual is determined using a posture recognition algorithm based on the key point information of the Siberian tiger individual.

4. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 3, characterized in that: The specific implementation process of the posture recognition algorithm is as follows: Calculate the mean value L of the horizontal coordinates of the key points 1, 2, and 3 of the head of the Siberian tiger in the image head , the mean value L of the horizontal coordinates of the hind limb key points 9 to 15 in the image tail , when L head <L tail When L head >L tail , its body is judged to be facing the right side.

5. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 1, characterized in that: In step three, first sample T frame images from the image sequence of the lateral area of ​​the Siberian tiger individual obtained in step two, and then resize them to the training size after letterbox operation to establish a Siberian tiger individual identification dataset; then construct and train a multi-frame Siberian tiger individual identification model based on convolutional neural network and graph neural network to obtain a Siberian tiger individual identification model.

6. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 5, characterized in that: The Siberian tiger individual recognition model comprises: an intra-frame feature extraction module based on a convolutional neural network, a multi-frame feature aggregation module based on a graph convolutional neural network, and a classification module; the intra-frame feature extraction module extracts features from each frame in an input sequence of images of the side area of ​​the Siberian tiger individual to obtain a corresponding feature map sequence; the obtained feature map is divided into blocks and a data map structure is constructed based on similarity; the multi-frame feature aggregation module interactively learns these feature blocks to obtain fused features; and the confidence information of the Siberian tiger individual is obtained through the classification module.

7. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 1, characterized in that: The specific implementation process of step 4 is as follows: S4.1: Use the wildlife camera numbers and time information obtained in step 1 to build a temporal and spatial information database of Siberian tiger individuals; S4.2: Define the set of wildlife cameras covering Siberian tiger individuals from nt-1 to n-1 weeks is the territory of the nth week, C represents the complete set of all cameras, namely C1C2,..., Indicates the camera number from 1 to N c The collection of N c represents the number of wildlife cameras; the probability P(n) of the migration of individual Siberian tiger territories in the nth week is counted; S4.3: Let the wildlife camera numbered j capture a Siberian tiger as event A j , the Siberian tiger with individual number i was photographed as event B i , the probability of a wild animal camera numbered j capturing an individual Siberian tiger is P(A j ) and the probability P(A) that a Siberian tiger with the number i is photographed by a wildlife camera with the number j j |B i ); S4.4: Based on the camera number and time information of the Siberian tiger when the individual was photographed, the probability of predicting the Siberian tiger individual is predicted by integrating the temporal and spatial information of the individual Siberian tiger; S4.5: judging the individual information of Siberian tigers based on the individual probability information of Siberian tigers calculated after integrating the spatiotemporal information; S4.6: Update the determined individual information of the Siberian tiger to the Siberian tiger individual spatiotemporal information database, and update the probability P(n) of the migration of the Siberian tiger individual territory every week.

8. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 7, characterized in that: The calculation formula for the probability P(n) of the migration of individual Siberian tiger territories in the nth week is: Among them, N t represents the number of Siberian tiger individuals; s represents the number of Siberian tiger individuals photographed in the territory; represents the number of times the Siberian tiger with individual number i was photographed in the territory in week n; C j (n) represents the number of Siberian tigers photographed by wildlife camera numbered j in week n.

9. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 7, characterized in that: The probability P(A) of the occurrence of an event in which the wildlife camera numbered j captures an individual Siberian tiger j ) and the probability P(A) that a Siberian tiger with the number i is photographed by a wildlife camera with the number j j |B i ) are calculated as follows: in, N represents the number of times the wildlife camera numbered j has photographed the Siberian tiger numbered i; t Indicates the number of Siberian tiger individuals.

10. The method for identifying Siberian tiger individuals by integrating spatiotemporal information according to claim 7, characterized in that: The calculation formula for the individual probability of Siberian tiger is: Where P(i) represents the probability that a Siberian tiger with number i is photographed by a wildlife camera with number j in week n; P(B i ) represents the probability of the Siberian tiger with individual number i appearing, i.e., the Siberian tiger individual confidence information predicted by the Siberian tiger individual recognition model in step 3; N t Represents the number of Siberian tiger individuals; The calculation formula for the individual information of the Siberian tiger is: Tiger id =argmax(P(i)),i∈(1,N t ) Among them, Tiger id represents the final predicted individual number of the Siberian tiger; P(i) represents the probability that the Siberian tiger individual belongs to number i; argmax(P(i)) represents the individual number corresponding to the predicted maximum probability value.

Images