A wild animal abnormal activity track real-time tracking method and system

Abstract

This application relates to the field of trajectory tracking technology, specifically to a method and system for real-time tracking of abnormal wildlife activity trajectories. The method includes: collecting location information of the wildlife; constructing predicted location information based on a state transition matrix; constructing a covariance matrix based on the predicted and location information, thereby determining a measurement noise covariance matrix; constructing a data acquisition window; calculating the deviation degree of each dimension based on the acquisition window; determining the anomaly weight by combining the difference between measured and predicted values; further determining the diagonal residual matrix; and then obtaining the Kalman gain matrix by combining the measurement noise covariance matrix; filtering and correcting the output signal based on the Kalman gain matrix; generating a heatmap from the signal; and monitoring the abnormal trajectory area. This application achieves more accurate tracking and monitoring.

Description

Technical Field

[0001] This application relates to the field of trajectory tracking technology, specifically to a method and system for real-time tracking of abnormal activity trajectories of wild animals. Background Technology

[0002] With the continuous destruction of animal habitats due to poaching and climate change, many wild animals, such as giant pandas and South China tigers, are gradually facing extinction. At the same time, some injured wild animals, or those involved in illegal trade, require timely treatment or assistance. Therefore, to protect and study these wild animals, many wildlife conservation bases and zoos have been built. These facilities not only provide refuge for endangered wild animals from habitat destruction, poaching, and climate change threats, but also delay species extinction through scientific breeding programs (such as captive breeding and genetic diversity management), and release healthy animal populations back into the wild, contributing to biodiversity. Real-time monitoring and analysis of the activity patterns of endangered animals at these facilities typically involves tracking their daily movements to assess the stability of their living conditions. When animals exhibit abnormal activity patterns due to special circumstances, deviating from their usual range, it may impact new areas and endanger protected wild animals.

[0003] To address changes in animal behavior promptly and support animal conservation and research in facilities like wildlife sanctuaries, real-time automatic tracking of abnormal animal movements using tracking devices is a crucial research direction. Currently, real-time tracking and monitoring of abnormal animal movements is primarily achieved using location tracking devices. However, animal movement can be affected by interference from their habitats, such as dense forests, canyons, or mountains within conservation areas, leading to signal obstruction, signal attenuation, and abrupt signal changes during location acquisition. Furthermore, there is a significant difference between the pre-set measurement noise covariance matrix of conventional Kalman filters and the actual interference noise encountered during signal acquisition, resulting in abnormal deviations in the acquired animal movement routes. Further solutions are needed to address these issues. Summary of the Invention

[0004] To address the technical problem of poor measurement accuracy caused by noise, this application provides a method and system for real-time tracking of abnormal wildlife activity trajectories. The specific technical solution adopted is as follows:

[0005] Firstly, this application proposes a method for real-time tracking of abnormal activity trajectories of wild animals, which includes the following steps:

[0006] Collect location information of wild animals;

[0007] The transition state matrix is ​​preset based on the speed changes of wild animals; the predicted position information for this acquisition is constructed based on the transition state matrix and the predicted position information collected in the previous acquisition; the position covariance matrix is ​​constructed based on the vector difference between the position information collected in this acquisition and the predicted position information; the measurement noise covariance matrix for this acquisition is constructed based on the position covariance matrix collected in this acquisition and the measurement noise covariance matrix collected in the previous acquisition.

[0008] A collection window is constructed for the first preset number of collections for each collection. The deviation of data in each dimension is calculated through the collection window. The outlier weight in each dimension is determined based on the difference and deviation between the measured value and the predicted value in each dimension. A diagonal residual matrix is ​​constructed for the outlier weights of all dimensions. The Kalman gain of the Kalman filter is reconstructed based on the measurement noise covariance matrix and the diagonal residual matrix to obtain the Kalman gain matrix.

[0009] The output signal is filtered and corrected based on the Kalman gain matrix, and then a heat map is generated from the signal to monitor abnormal trajectory areas.

[0010] In the aforementioned scheme, this application constructs an adaptive Kalman gain. Considering that when tracking abnormal animal activity trajectories in nature reserves, the positioning signal may attenuate or experience abrupt signal changes when the animal moves into dense forests or other areas where occlusion occurs, conventional Kalman filtering uses a fixed measurement noise covariance matrix. When animal activity is affected by environmental occlusion, leading to signal attenuation or abrupt changes in the tracking device, the preset measurement noise covariance matrix differs significantly from the actual interference noise encountered during signal acquisition, potentially causing large trajectory deviations. This application constructs an adaptive measurement noise covariance matrix to reduce the weight of the gain on the measured data when signal attenuation occurs due to occlusion. Furthermore, it uses a diagonal residual matrix to suppress abrupt outliers when single-dimensional variable changes occur, preventing the algorithm from being affected by single-dimensional abrupt changes when correcting the signal, thus avoiding impacts on all dimensions. The optimized Kalman filter then achieves more accurate tracking and monitoring through precise correction.

[0011] In one embodiment, the dimensions of the location information include longitude, latitude, longitude velocity, latitude velocity, longitude acceleration, and latitude acceleration.

[0012] In one embodiment, the method of pre-setting the transition state matrix based on the speed changes of wild animals is as follows:

[0013] Select the first n data points from each location data collection session;

[0014] When the total velocity of the first n data points continuously increases, it is set to uniform acceleration mode; when the total velocity of the first n data points continuously decreases, it is set to uniform deceleration mode; when the total velocity of the first n data points remains unchanged, it is set to uniform speed mode. Then, a state transition matrix is ​​constructed through a motion model. The total velocity is the sum of the longitude velocity and the latitude velocity.

[0015] In one embodiment, the location covariance matrix is ​​constructed as follows:

[0016] , This represents the position vector of the position information collected in the kth instance. This represents the position vector of the predicted position information collected in the k-th acquisition. This represents the matrix transpose operation. This represents the position covariance matrix of the k-th acquisition.

[0017] The position vector is a vector composed of all dimensions of the position information.

[0018] In one embodiment, the method for constructing the measurement noise covariance matrix is ​​as follows:

[0019] , Represents the smoothing factor. This represents the measurement noise covariance matrix of the (k-1)th acquisition. This represents the position covariance matrix of the k-th data acquisition. Represents a diagonal matrix function. Let represent the measurement noise covariance matrix of the k-th acquisition.

[0020] In one embodiment, the method for calculating the deviation of each dimension of data through the acquisition window is as follows:

[0021] For each dimension of data within the acquisition window, after fitting the data for each dimension, calculate the residual between the fitted value and the measured value, and use the standard deviation of all residuals as the degree of deviation of the data for each dimension.

[0022] In one embodiment, the anomaly weight is positively correlated with the difference between the measured value and the predicted value, and negatively correlated with the degree of deviation; the predicted value for each dimension is the value of the dimension corresponding to the location vector of the predicted location information.

[0023] In one embodiment, the diagonal residual matrix is ​​a diagonal matrix formed by the outlier weights of all dimensions in the order of their position vectors.

[0024] In one embodiment, the Kalman gain matrix is ​​expressed as follows:

[0025] , Let represent the prior error covariance matrix of the u-th acquisition. Represents the latitude and longitude observation matrix. This represents the transpose of the latitude and longitude observation matrix. Let represent the diagonal residual matrix of the u-th acquisition. Let represent the measurement noise covariance matrix of the u-th acquisition. This represents the Kalman gain matrix for the u-th acquisition; the latitude and longitude observation matrix has longitude and latitude values ​​set to 1, and the other dimensions set to 0.

[0026] Secondly, embodiments of this application also provide a real-time tracking system for abnormal activity trajectories of wild animals, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described methods for real-time tracking of abnormal activity trajectories of wild animals.

[0027] The beneficial effects of this application are as follows:

[0028] This application constructs an adaptive Kalman gain to address the issue of signal attenuation and abrupt changes in location signals when tracking abnormal animal movements in nature reserves, particularly when animals move into dense forests or other areas with occlusion. Conventional Kalman filtering uses a fixed measurement noise covariance matrix, which is susceptible to signal attenuation and abrupt changes due to environmental occlusion during animal activity. The preset measurement noise covariance matrix differs significantly from the actual noise interference encountered during signal acquisition, potentially leading to large trajectory deviations. This application constructs an adaptive measurement noise covariance matrix to reduce the weight of the gain on the measured data when signal attenuation occurs due to occlusion. Furthermore, it uses a diagonal residual matrix to suppress outliers caused by single-dimensional variable abrupt changes, preventing the algorithm from being affected by single-dimensional abrupt changes in all dimensions when correcting the signal. The optimized Kalman filter then achieves more accurate tracking and monitoring through precise correction. Attached Figure Description

[0029] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart illustrating a method for real-time tracking of abnormal activity trajectories of wild animals, provided as an embodiment of this application. Detailed Implementation

[0031] To further illustrate the technical means and effects adopted by this application to achieve the intended purpose of the invention, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a real-time tracking method and system for abnormal activity trajectories of wild animals proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0033] An embodiment of a method and system for real-time tracking of abnormal activity trajectories of wild animals:

[0034] The following description, in conjunction with the accompanying drawings, details the specific scheme of the real-time tracking method and system for abnormal wildlife activity provided in this application.

[0035] Please see Figure 1 The diagram illustrates a flowchart of a real-time tracking method for abnormal wildlife activity trajectories according to an embodiment of this application. The method includes the following steps:

[0036] Step S001: Collect location information of wild animals.

[0037] In real-time automatic tracking of abnormal animal activity trajectories in wildlife conservation bases, the scenarios typically involve complex features, such as dense forests, canyons, and mountains. Therefore, when locating and tracking animal movement trajectories in wildlife conservation bases, tracking necklaces are usually used to obtain real-time location information of the animals. In this embodiment, wild animals in the wildlife conservation base are selected as the targets for tracking abnormal activity trajectories. By configuring an nRF9160-based wildlife tracker on the wild animals, the GPS location signal of the wild animals is obtained in real time through the tracker, i.e., location information is acquired. In this embodiment, the location information is represented by longitude, latitude, longitude velocity, latitude velocity, longitude acceleration, and latitude acceleration.

[0038] At this point, the location information for each sample has been obtained.

[0039] Step S002: Construct predicted location information based on the state transition matrix, construct a covariance matrix based on the predicted location information and the location information, and thus determine the measurement noise covariance matrix.

[0040] In nature reserves, dense forests, canyons, and mountains often obstruct the view, and weather changes such as heavy rain can cause complex noise interference when acquiring signals. To ensure the stability and effectiveness of GPS positioning signals acquired when wild animals are moving within nature reserves, Kalman filtering is used to predict and correct the acquired location signals.

[0041] When using conventional Kalman filtering to predict and correct the location signals of wildlife movement trajectories in nature reserves, it's important to consider the real-time changes in noise levels caused by the natural environment as wildlife moves through the reserve. For example, sudden noise spikes occur during heavy rain, and obstructions such as leaves and tree trunks interfere with movement through dense forests. These factors affect the collected location information, leading to significant deviations from the predicted location information.

[0042] The predicted position information is obtained by multiplying the state transition matrix by the predicted position information from the previous time step. The state transition matrix is ​​determined by the motion model. The motion model is constructed by selecting the velocity changes from the n signals acquired before each acquisition time. To increase prediction accuracy, the value of n is set to [4,7]. In this embodiment, n is taken as an empirical value of 5. The motion model is set according to the velocity data changes. When the velocity increases sequentially over time (i.e., the velocity at each time step is greater than the velocity at the previous time step), it is set to a uniform acceleration mode; when the velocity decreases sequentially over time (i.e., the velocity at each time step is less than the velocity at the previous time step), it is set to a uniform deceleration mode; otherwise, it is set to a uniform speed mode. The state transition matrix is ​​obtained based on different modes. The predicted position information at the first time step is the position information at the first time step.

[0043] For each sample, its location information is used to form a location vector, which is a vector composed of all dimensions of the location information. The predicted location information is also used to form a location vector. The location covariance matrix is ​​constructed based on the difference between the two location vectors and their transpose.

[0044] Preferably, in this embodiment, the method for constructing the location covariance matrix is ​​as follows:

[0045] , This represents the position vector of the position information collected in the kth instance. This represents the position vector of the predicted position information collected in the k-th acquisition. This represents the matrix transpose operation. Let represent the position covariance matrix of the k-th acquisition. The diagonal data of the acquired covariance matrix represents the squared deviation between the data of the corresponding dimension's position information and the predicted position information, indicating the impact of interference noise on the signal.

[0046] When wildlife tracked by a track is affected by environmental factors (such as signal dispersion caused by rain when an animal is running continuously in the rain), the acquired location information deviates significantly from the predicted location information, and the diagonal value of the covariance matrix increases significantly. However, when the animal moves under normal conditions without interference, the diagonal value of the acquired covariance matrix is ​​relatively small.

[0047] Conventional Kalman filtering typically uses a preset measurement noise covariance matrix to predict and correct signals. However, in situations such as sudden rainstorms or dense forest obstructions that cause signal scattering changes, there is a significant difference between the preset measurement noise covariance matrix and the interference noise encountered during the actual location information acquisition, which leads to a decrease in filtering effectiveness.

[0048] Therefore, the measurement noise covariance matrix of the current acquisition is obtained by weighting the position covariance matrix of each acquisition with the measurement noise covariance matrix of the previous acquisition.

[0049] Preferably, in this embodiment, the expression for the measurement noise covariance matrix is:

[0050] , Represents the smoothing factor. This represents the measurement noise covariance matrix of the (k-1)th acquisition. This represents the position covariance matrix of the k-th data acquisition. Represents a diagonal matrix function. Let represent the measurement noise covariance matrix of the k-th acquisition.

[0051] The diagonal matrix function represents a matrix in which all elements except those on the main diagonal are set to 0; the measurement noise covariance matrix acquired in the first acquisition is determined by the error setting of the measurement equipment itself.

[0052] When the location signal of a wild animal being tracked is affected by environmental factors and produces errors, the diagonal function of the location covariance matrix increases significantly, and the updated measurement noise covariance matrix also increases accordingly. However, when the animal resumes normal movement without interference, the obtained residual covariance term takes a very small value, and the noise covariance matrix gradually stabilizes with each update.

[0053] Thus, the measurement noise covariance matrix was obtained.

[0054] Step S003: Construct a data acquisition window, calculate the degree of deviation for each dimension based on the data acquisition window, determine the anomaly weight by combining the difference between the measured value and the predicted value, then determine the diagonal residual matrix, and finally obtain the Kalman gain matrix by combining the measurement noise covariance matrix.

[0055] Considering that the location information of animals is obtained by using tracking devices, the location signal is affected by obstructions from all directions when the animal crosses the dense forest. Since the latitude, longitude, speed and acceleration data obtained by the tracking device are usually obtained by multiple sensors, when acquiring multi-dimensional data and collecting signals synchronously, latitude and longitude signals may be lost or jump due to the animal being blocked or electromagnetic interference during thunderstorms, resulting in instantaneous abnormal values, which in turn leads to relatively lower reliability of the measured data.

[0056] Based on the above analysis, in order to address the possible abnormal values, the weights of the corresponding dimensions of the measurement noise covariance matrix are calculated. For each acquisition, the first preset number of acquisitions M are used as a sliding window. To ensure that the acquired data is not affected too much by the long time interval, the value of M is set to [3,6]. In this embodiment, the value of M is 5.

[0057] For cases where the number of data collections is less than M, the data is supplemented by using the mean to construct a noise weight index, which is used to characterize the weight increase of the corresponding dimension of the measurement noise covariance matrix when outliers occur in the corresponding dimension.

[0058] In Kalman filtering, an observation matrix is ​​set up, where each dimension corresponds to an observation matrix. The observation matrix for each dimension has a value of 1 for that dimension, and the data for the other dimensions is 0.

[0059] For each sliding window, the data for each dimension is fitted and the residuals are calculated. Then, the standard deviation of the residuals is calculated, which represents the degree of deviation between the measured value and the predicted value.

[0060] The anomaly weight in each dimension is determined by comparing the difference and deviation between the measured value and the predicted value in each dimension.

[0061] The abnormal weights are positively correlated with the difference between the measured and predicted values, and negatively correlated with the degree of deviation.

[0062] It should be noted that positive correlation means that when one variable increases, the other variable also increases, and the two variables change in the same direction. When one variable changes from large to small or from small to large, the other variable also changes from large to small or from small to large. The specific relationship is determined by the actual application, and this application does not impose any special restrictions.

[0063] It should be noted that negative correlation means that when one variable increases, the other variable decreases accordingly, and the two variables change in opposite directions. When one variable changes from large to small or from small to large, the other variable also changes from small to large or from large to small. The specific relationship is determined by practical application, and this application does not impose any special restrictions.

[0064] Preferably, in this embodiment, the expression for the anomaly weight is:

[0065] , This represents the observation matrix of dimension c. This represents the data of dimension c representing the location information collected in the kth instance. This represents the position vector of the predicted position information collected in the k-th acquisition. This represents the standard deviation of the residuals of the data in dimension c within the sliding window corresponding to the k-th data collection. This represents a very small positive number, used to ensure that the denominator is not zero. Represents the normalization function. This represents the abnormal weight of dimension c in the k-th collection.

[0066] The anomaly weights of all dimensions collected in the kth sampling are arranged in the order of the position vectors to form a diagonal matrix, which is denoted as the diagonal residual matrix of the noise weights.

[0067] The weights calculated for each dimension are formed into a diagonal matrix. When outliers such as jumps or missing data occur in the corresponding dimension, The calculated value is much higher than the standard deviation of the residuals within the sliding window, resulting in a higher D value and a larger weight; however, when calculating data under normal, interference-free conditions... The smaller the calculated value, the smaller the obtained D value, and the smaller the obtained weight.

[0068] The Kalman gain of the Kalman filter is reconstructed based on the measurement noise covariance matrix and the diagonal residual matrix to obtain the Kalman gain matrix.

[0069] Preferably, the expression for the Kalman gain matrix is:

[0070] , Let represent the prior error covariance matrix of the u-th acquisition. Represents the latitude and longitude observation matrix. This represents the transpose of the latitude and longitude observation matrix. Let represent the diagonal residual matrix of the u-th acquisition. Let represent the measurement noise covariance matrix of the u-th acquisition. Let represent the Kalman gain matrix for the u-th acquisition.

[0071] The latitude and longitude observation matrix has longitude and latitude values ​​set to 1, and other dimensions set to 0. The prior error covariance matrix is ​​obtained by adding the process noise covariance matrix to the state transition matrix and the error covariance matrix of the previous time step. The process noise covariance matrix is ​​derived through motion model derivation. The first error covariance matrix is ​​obtained through the equipment's own error.

[0072] Thus, the Kalman gain matrix of the Kalman filter has been obtained.

[0073] Step S004: Filter and correct the output signal based on the Kalman gain matrix, generate a heatmap from the signal, and then monitor the abnormal trajectory area.

[0074] The location signals of animals in the wildlife sanctuary, obtained through a positioning device, are filtered and corrected using a Kalman filter optimized with a Kalman gain matrix. The filtered location signal data is then output. Kalman filtering is a well-known technique, and its specific operation will not be elaborated upon further. The designated activity range of the wildlife sanctuary is obtained by summarizing the acquired location signal data into the corresponding area image. A one-week historical movement trajectory of the animal is taken as a period. After summarizing the animal's trajectory routes within this period into the local area image, a heatmap is generated using Folium based on the trajectory routes. The heatmap reflects the animal's location preferences within the wildlife sanctuary. The area with the fewest location information in the period is identified as an abnormal trajectory area (e.g., tigers usually patrol and hunt prey in their own territory and generally do not venture into other tiger territories. When a target tiger ventures into another tiger's territory, it indicates that the target tiger is exhibiting abnormal activity, which may trigger a fight between different tigers). When an animal appears in a corresponding abnormal trajectory area, the positioning device tracks the animal's abnormal activity trajectory in real time, and the abnormal activity trajectory tracking data is fed back as a real-time warning, thus completing the real-time tracking of the animal's abnormal activity trajectory.

[0075] This completes the tracking of the animal's abnormal activity.

[0076] Based on the same inventive concept as the above methods, embodiments of the present invention also provide a real-time tracking system for abnormal activity trajectories of wild animals, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above methods for real-time tracking of abnormal activity trajectories of wild animals.

[0077] It should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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 of the technical features. Such 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 this application, and should all be included within the protection scope of this application.

[0078] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for real-time tracking of abnormal activity trajectory of wild animals, characterized in that, The method includes the following steps: Collect location information of wild animals; The transition state matrix is ​​preset based on the speed changes of wild animals; the predicted position information at the current moment is constructed based on the transition state matrix and the predicted position information at the previous moment; the position covariance matrix is ​​constructed based on the vector difference between the current position information and the predicted position information; the measurement noise covariance matrix at the current moment is constructed based on the position covariance matrix at the current moment and the measurement noise covariance matrix at the previous moment. A collection window is constructed for a preset number of collections at the current moment. The deviation of data in each dimension is calculated through the collection window. The outlier weight in each dimension is determined based on the difference between the measured value and the predicted value of each dimension, as well as the above deviation. A diagonal residual matrix is ​​constructed for the outlier weights of all dimensions. The Kalman gain of the Kalman filter is reconstructed based on the measurement noise covariance matrix and the diagonal residual matrix to obtain the Kalman gain matrix. The output signal is filtered and corrected based on the Kalman gain matrix, and then a heat map is generated from the signal to monitor abnormal trajectory areas. The method for constructing the measurement noise covariance matrix is ​​as follows: , denotes a smoothing factor, denotes the measurement noise covariance matrix of the (k-1)th acquisition, denotes the position covariance matrix of the kth acquisition, denotes a diagonal matrix function, denotes the measurement noise covariance matrix of the kth acquisition; The expression for the Kalman gain matrix is: , Let represent the prior error covariance matrix of the u-th acquisition. Represents the latitude and longitude observation matrix. This represents the transpose of the latitude and longitude observation matrix. Let represent the diagonal residual matrix of the u-th acquisition. Let represent the measurement noise covariance matrix of the u-th acquisition. This represents the Kalman gain matrix for the u-th acquisition; the latitude and longitude observation matrix has longitude and latitude values ​​set to 1, and the other dimensions set to 0.

2. The method for real-time tracking of abnormal activity trajectories of wild animals as described in claim 1, characterized in that, The dimensions of the location information include longitude, latitude, longitude velocity, latitude velocity, longitude acceleration, and latitude acceleration.

3. The method for real-time tracking of abnormal activity trajectories of wild animals as described in claim 1, characterized in that, The method for pre-setting a transition state matrix based on the speed changes of wild animals is as follows: Select the first n data points from each location data collection session; When the total velocity of the first n data points continuously increases, it is set to uniform acceleration mode; when the total velocity of the first n data points continuously decreases, it is set to uniform deceleration mode; when the total velocity of the first n data points remains unchanged, it is set to uniform speed mode. Then, a state transition matrix is ​​constructed through a motion model. The total velocity is the sum of the longitude velocity and the latitude velocity.

4. The method for real-time tracking of abnormal activity trajectories of wild animals as described in claim 1, characterized in that, The method for constructing the location covariance matrix is ​​as follows: , This represents the position vector of the position information collected in the kth instance. This represents the position vector of the predicted position information collected in the k-th acquisition. This represents the matrix transpose operation. This represents the position covariance matrix of the k-th acquisition. The position vector is a vector composed of all dimensions of the position information.

5. The method for real-time tracking of abnormal activity trajectories of wild animals as described in claim 1, characterized in that, The method for calculating the deviation of each dimension of data through the acquisition window is as follows: For each dimension of data within the acquisition window, after fitting the data for each dimension, calculate the residual between the fitted value and the measured value, and use the standard deviation of all residuals as the degree of deviation of the data for each dimension.

6. The method for real-time tracking of abnormal activity trajectories of wild animals as described in claim 1, characterized in that, The anomaly weights are positively correlated with the difference between the measured and predicted values, and negatively correlated with the degree of deviation; the predicted value for each dimension is the value of the dimension corresponding to the predicted location vector.

7. The method for real-time tracking of abnormal activity trajectories of wild animals as described in claim 1, characterized in that, The diagonal residual matrix is ​​a diagonal matrix formed by the anomaly weights of all dimensions in the order of their position vectors.

8. A real-time tracking system for abnormal wildlife activity, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method for real-time tracking of abnormal activity trajectories of wild animals as described in any one of claims 1-7.

Images