Multi-source array animal intelligent monitoring and driving method and system based on target recognition

Abstract

The invention discloses a multi-source array animal intelligent monitoring and driving method and system based on target recognition, and relates to the technical field of animal prevention and control, and the method comprises the steps: based on the collection data of a multi-source sensor array, extracting the distribution behavior parameters of animal individuals, obtaining an individual behavior feature data set, and obtaining an individual behavior feature data set; according to the number of back-and-forth movement times of the animal individuals in the monitoring boundary area, the single staying duration and the ratio of the entering depth to the exiting frequency, an animal heuristic behavior judgment result is generated; according to the animal heuristic behavior judgment result and the historical driving event record, calculating adaptability parameters of the group to acousto-optic driving stimulation, setting an adaptability critical threshold judgment standard, and generating an animal group driving adaptability evaluation result; and generating a driving control signal containing the starting instruction and the adjusting parameter. The method has the advantages of animal heuristic behavior recognition and driving adaptability evaluation, and solves the problem of poor long-term prevention and control effect in the prior art.

Description

Technical Field

[0001] This invention relates to the field of animal control technology, and more specifically, to a multi-source array intelligent animal monitoring and repelling method and system based on target recognition. Background Technology

[0002] With increased efforts in ecological environmental protection and the recovery of wild animal populations, the damage caused by animals invading human activity areas is becoming increasingly serious, urgently requiring intelligent animal control technologies. Taking wild boars as an example, their populations can expand rapidly and frequently invade agricultural areas, causing huge economic losses to crops. Wild boars possess strong intelligence and learning abilities, enabling them to quickly adapt to environmental changes and develop complex group behavior patterns, rendering traditional physical barriers and manual deterrence methods ineffective. With the development of artificial intelligence and multi-sensor fusion technologies, automated animal monitoring and deterrence technologies based on computer vision and intelligent algorithms are gradually becoming an important development direction in this field, providing a new technological path for solving animal control problems.

[0003] Existing animal monitoring and deterrence technologies primarily employ a single sensor node for target detection. This involves capturing image data via a camera, using deep learning models to identify animal targets, and triggering a pre-defined deterrence program upon detecting an intrusion. These technologies typically rely on convolutional neural networks for target classification, combined with motion detection algorithms to determine the animal's activity level. When trigger conditions are met, an audio-visual deterrence device is activated to drive the animal away. While this direct triggering method offers rapid response, it lacks in-depth analysis of animal behavior, often leading to false triggers and wasted resources.

[0004] However, existing animal monitoring and deterrence technologies have significant shortcomings. First, they lack the ability to identify intelligent behavioral patterns within animal groups. Some animal groups often send individuals to probe ahead, exhibiting complex counter-surveillance behaviors, which current technologies cannot identify. This makes it impossible to accurately determine group size and behavioral intentions, thus affecting the formulation and effectiveness of deterrence strategies. Second, they do not consider the adaptability of animals to deterrence measures. After experiencing the same deterrence pattern multiple times, animal groups gradually adapt and may even completely ignore deterrence measures. Current technologies lack an assessment mechanism for this adaptation, making it impossible to determine whether the effectiveness of the current deterrence strategy is decreasing. This adaptive learning is a natural survival instinct for animals, and current technologies ignore this crucial factor, resulting in poor long-term protective effects. These two problems lead to a gradual weakening of the effectiveness of existing technologies in long-term applications. When facing animals with adaptive learning behaviors (such as wild boars), the deterrence effect often diminishes, and the misjudgment rate increases, making the overall protective effect insufficient to meet practical application needs.

[0005] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention

[0006] To address the problems in related technologies, this invention proposes a multi-source array animal intelligent monitoring and repelling method and system based on target recognition. It has the advantages of animal probing behavior recognition and repelling adaptability assessment, thereby solving the problems of existing technologies that cannot predict animal intrusion intentions and the gradual decrease in repelling effect.

[0007] Therefore, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a multi-source array intelligent monitoring and repelling method for animals based on target recognition is provided, the method comprising: S1. Based on the data collected by the multi-source sensor array, extract the distribution behavior parameters of individual animals to obtain an individual behavior feature dataset, and generate the animal exploration behavior judgment result according to the number of times the individual animal moves back and forth in the monitoring boundary area, the duration of a single stay, and the ratio of entry depth to exit frequency. S2. Based on the results of animal trial behavior and historical driving event records, the adaptive parameters of the group to the sound and light driving stimuli are calculated using the response threshold enhancement algorithm, and the adaptive critical threshold judgment criteria are set to generate the animal group driving adaptation assessment results. S3. Based on the results of animal trial behavior assessment and animal group drive-off adaptability assessment, generate a drive-off control signal containing activation instructions and adjustment parameters.

[0008] Furthermore, the collected data includes visual, acoustic, and thermal imaging data; the distributed behavioral parameters include position coordinate sequences, rate of change of movement speed, and dwell time distribution; based on the collected data from the multi-source sensor array, the distributed behavioral parameters of individual animals are extracted to obtain an individual behavioral feature dataset, including: obtaining the pixel coordinates of individual animals in images through visual data, converting the pixel coordinates into visual position coordinates according to a preset camera calibration matrix, and recording spatial coordinate changes in a time series to form a position coordinate sequence; identifying the target animal's sound based on acoustic data, calculating the time difference of the target animal's sound reaching each acoustic sensor to determine the sound source location, associating and matching the sound source location with the position coordinate sequence, and solving for the rate of change of movement speed; marking individual animal thermal targets based on thermal imaging data, statistically analyzing the duration of the centroid coordinates of individual animal thermal targets to determine dwell events, calculating the dwell time distribution of individual animals within the monitoring area, and combining the position coordinate sequence and the rate of change of movement speed to establish an individual behavioral feature dataset.

[0009] Furthermore, based on the number of times an animal moves back and forth within the monitoring boundary area, the duration of each stay, and the ratio of entry depth to exit frequency, the animal's exploratory behavior is determined, including: setting the coordinates of the monitoring boundary line, statistically analyzing the entry and exit events of an animal crossing the monitoring boundary line in the position coordinate sequence, and accumulating the number of back and forth movements; determining the stay events of an animal within the monitoring area based on the stay time distribution, extracting the start and end times of each stay event, and calculating the time difference to obtain the duration of each stay; using the position coordinate sequence to calculate the farthest distance from the monitoring boundary line after the animal enters the monitoring area as the entry depth, combining the rate of change of movement speed to statistically analyze the exit frequency within a preset time period, and calculating the ratio of entry depth to exit frequency; when the number of back and forth movements exceeds a first preset threshold, the duration of each stay exceeds a second preset threshold, and the ratio of entry depth to exit frequency exceeds a third preset threshold, it is determined as animal exploratory behavior and output.

[0010] Furthermore, the adaptive parameters of a group to audiovisual repulsion stimuli include the initial response threshold, the current response threshold, and the threshold change rate. Based on the animal trial behavior assessment results and historical repulsion event records, the adaptive parameters of a group to audiovisual repulsion stimuli are calculated using a response threshold enhancement algorithm, including: establishing a historical repulsion event database, recording the time, location, type of repulsion stimulus, stimulus intensity, and animal group response behavior for each repulsion event; determining the initial response threshold based on the stimulus intensity when the animal group begins to flee in the first repulsion event; and constructing a probabilistic model based on the response threshold, according to the animal trial behavior... The behavioral assessment results and historical driving event database are used to analyze the animal population's adaptation to driving stimuli. The trend of animal population response to the same type of stimuli in historical driving events is analyzed, and the minimum stimulus intensity that causes the population to flee in each driving event is calculated to obtain the current response threshold. The ratio of the current response threshold to the initial response threshold is calculated to obtain the threshold change factor, and the growth rate of the threshold change factor per unit time is calculated to obtain the threshold change rate. The threshold change factor is multiplied by a preset reference coefficient, and the threshold change rate is multiplied by a preset weight to obtain the population's adaptation parameters to sound and light driving stimuli.

[0011] According to another aspect of the present invention, a multi-source array intelligent animal monitoring and deterrence system based on target recognition is also provided, the multi-source array intelligent animal monitoring and deterrence system based on target recognition includes: The individual exploration module is used to extract the distribution behavior parameters of individual animals based on the data collected by the multi-source sensor array, obtain the individual behavior feature dataset, and generate the animal exploration behavior judgment result based on the number of times the individual animal moves back and forth in the monitoring boundary area, the duration of a single stay, and the ratio of entry depth to exit frequency. The group adaptation module is used to calculate the group's adaptation parameters to the sound and light driving stimuli based on the results of animal trial behavior judgment and historical driving event records, using the response threshold enhancement algorithm, and to set the adaptation critical threshold judgment criteria to generate animal group driving adaptation assessment results. The regulation and control module is used to generate a deportation control signal containing activation commands and regulation parameters based on the results of animal trial behavior judgment and animal group deportation adaptability assessment.

[0012] The beneficial effects of this invention are as follows: (1) This invention solves the technical problem of difficulty in identifying individual animal probing behavior by working in concert with a multi-source sensor array; the system can capture and analyze the subtle behavioral characteristics of animals moving back and forth and observing in the boundary area, thereby identifying the probing behavior of the advance individuals before a large-scale invasion, turning the protective measures from passive response to active prevention, and greatly improving the timeliness and accuracy of prevention and control; compared with the existing single sensor triggering mechanism, this invention can detect the anti-reconnaissance behavior pattern of the advance individuals in time before a large-scale invasion, enabling managers to deploy targeted protective measures in advance, and significantly improving the initiative and effectiveness of animal invasion prevention and control.

[0013] (2) This invention innovatively establishes an adaptive assessment system for animal groups to drive-away measures. By recording and analyzing the changing trends of animal responses in historical drive-away events, it can scientifically determine whether the current drive-away strategy is failing. When the drive-away effect is detected to be decreasing, the type, intensity, and timing of the sound and light stimuli are automatically adjusted, effectively overcoming the adaptation problem of animals to a single drive-away mode. Practical application shows that compared with traditional fixed-mode drive-away, this scheme significantly extends the duration of drive-away effectiveness, providing a more reliable long-term protection solution for agricultural production and ecological protection. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating a multi-source array-based intelligent animal monitoring and repelling method based on target recognition according to an embodiment of the present invention. Figure 2 This is a specific implementation diagram of generating animal probing behavior judgment results in a multi-source array animal intelligent monitoring and driving method based on target recognition according to an embodiment of the present invention; Figure 3This is a schematic diagram of a multi-source array intelligent animal monitoring and driving system based on target recognition according to an embodiment of the present invention. Detailed Implementation

[0016] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0017] According to embodiments of the present invention, a multi-source array animal intelligent monitoring and driving method and system based on target recognition is provided.

[0018] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a multi-source array animal intelligent monitoring and repelling method based on target recognition is provided, the multi-source array animal intelligent monitoring and repelling method based on target recognition includes: S1. Based on the data collected by the multi-source sensor array, extract the distribution behavior parameters of individual animals to obtain an individual behavior feature dataset, and generate the animal exploration behavior judgment result according to the number of times the individual animal moves back and forth in the monitoring boundary area, the duration of a single stay, and the ratio of entry depth to exit frequency. S2. Based on the results of animal trial behavior and historical driving event records, the adaptive parameters of the group to the sound and light driving stimuli are calculated using the response threshold enhancement algorithm, and the adaptive critical threshold judgment criteria are set to generate the animal group driving adaptation assessment results. S3. Based on the results of animal trial behavior assessment and animal group drive-off adaptability assessment, generate a drive-off control signal containing activation instructions and adjustment parameters.

[0019] In one embodiment, the collected data includes visual, acoustic, and thermal imaging data; the distribution behavior parameters include position coordinate sequences, rate of change of movement speed, and residence time distribution. like Figure 2 As shown, in step S1, based on the data collected by the multi-source sensor array, the distribution behavior parameters of individual animals are extracted to obtain the individual behavior feature dataset, which includes: The pixel coordinates of individual animals in the image are obtained through visual data. The pixel coordinates are converted into visual position coordinates according to the preset camera calibration matrix, and the spatial coordinate changes are recorded in time series to form a position coordinate sequence. Based on acoustic data, the sound of the target animal is identified, and the time difference of the sound reaching each acoustic sensor is calculated to determine the location of the sound source. The sound source location is then correlated and matched with the location coordinate sequence to solve for the rate of change of movement speed. Animal thermal targets are marked based on thermal imaging data. The duration of the centroid coordinates of the animal thermal targets is counted to determine dwell events. The dwell time distribution of animal individuals in the monitoring area is calculated. Combined with the position coordinate sequence and the rate of change of movement speed, an individual behavior feature dataset is established.

[0020] In one embodiment, obtaining the pixel coordinates of an individual animal in an image through visual data, converting the pixel coordinates into visual position coordinates according to a preset camera calibration matrix, and recording the spatial coordinate changes in a time series to form a position coordinate sequence includes: Based on image data collected by different visual sensors within the monitoring area (the locations of the preset multiple sensors will be explained in subsequent embodiments), inter-frame difference is performed on each visual sensor image to extract the moving target region; The extracted moving target region is matched and identified using an animal body feature template (explained in subsequent embodiments), the individual animal contour region is marked, and the pixel coordinates of the contour centroid are extracted; The pixel coordinates extracted from each visual sensor are converted into visual position coordinates in a unified three-dimensional coordinate system through a pre-calibrated camera intrinsic and extrinsic parameter matrix. Spatial coordinates obtained from different visual sensors at the same time point are weighted and fused to eliminate viewpoint errors and form a time-stamped sequence of location coordinates.

[0021] In one embodiment, the sound of a target animal is identified based on acoustic data, and the time difference between the arrival of the sound at each acoustic sensor is calculated to determine the location of the sound source. The sound source location is then correlated and matched with the location coordinate sequence to solve for the rate of change of movement speed, including: Based on an acoustic sensor array (the arrangement of multiple acoustic sensors to form an acoustic array will be explained in subsequent embodiments), sound signals generated by animal activities are collected and preprocessed to eliminate background noise; Extract the time-frequency features of the sound signal and identify the target animal's sound based on a pre-set animal sound feature database; The time difference between the arrival of the target animal's sound at each acoustic sensor is calculated. Combined with the spatial distribution parameters of the acoustic sensors, the spatial coordinates of the sound source are calculated using a sound source localization algorithm (explained in subsequent embodiments). The spatial coordinates of the sound source are associated and matched with the visual position coordinates at the corresponding time points. The displacement distance between adjacent sampling time points is calculated by the position coordinate sequence within the time window and divided by the time interval to obtain the rate of change of the animal's movement speed in each time period.

[0022] In one embodiment, marking individual animal thermal targets based on thermal imaging data, statistically analyzing the duration of the centroid coordinates of the individual animal thermal targets to determine dwell events, and calculating the dwell time distribution of individual animals within the monitoring area includes: Based on the collected thermal imaging data of the monitoring area, upper and lower limits of temperature thresholds are set according to the body temperature characteristics of the target animal. The thermal imaging data is binarized and thermal signal areas that meet the temperature threshold range are extracted. Connectivity analysis is performed on the thermal signal region to remove noise regions smaller than a preset area threshold, and thermal signal regions that meet the size condition are marked as thermal targets of individual animals. Calculate the centroid coordinates of the thermal target of an individual animal and establish a mapping relationship between the centroid coordinates and the position coordinate sequence; Based on the mapping relationship, the duration of the thermal target of an individual animal in a continuous frame sequence is statistically analyzed, and the corresponding three-dimensional coordinates are recorded. When the centroid coordinates are within a preset spatial range and the duration exceeds a preset time threshold, it is determined as a dwell event. The total duration of dwell time for each three-dimensional coordinate is accumulated to form a heat map of the dwell time distribution of individual animals within the monitoring area, and the dwell time distribution is extracted.

[0023] Specifically, the multi-source sensor array in this embodiment includes multiple visual sensors, acoustic sensors, and thermal imaging sensors arranged around the boundary of the monitoring area. The visual sensors are waterproof high-definition network cameras, mounted on 3-meter-high pillars at 5-meter intervals along the monitoring boundary line, with a field of view of 120 degrees and ensuring a 30% overlap between adjacent cameras. The acoustic sensors are omnidirectional microphone arrays, each group containing six microphones in a hexagonal layout, with a 2-meter spacing between adjacent microphones, arranged every 30 meters along the boundary of the monitoring area. The thermal imaging sensors are uncooled long-wave infrared thermal imagers, installed at key monitoring points to cover the blind spots of the visual sensors. All sensors are connected to a local computing device via wired or wireless networks, employing a unified time synchronization mechanism to ensure accurate alignment of data timestamps, enabling collaborative acquisition and processing of multi-source data.

[0024] Specifically, in this embodiment, the animal body feature template is a target animal morphological feature recognition model constructed using deep learning methods. For common invasive animals such as wild boars, 1000 sample images each containing different walking postures and body sizes were collected. The YOLOv5 algorithm framework was used to train the target recognition model, which achieved a detection accuracy of over 95%. The sound source localization algorithm adopts a hyperbolic intersection localization method based on the time difference of arrival (TDOA) principle, combined with a Bayesian filtering algorithm to eliminate environmental noise interference. In thermal imaging data processing, based on the normal body temperature range of wild boars (38℃-39.5℃), the upper limit of the temperature threshold is set to 40℃ and the lower limit is set to 37℃ to adapt to the surface temperature characteristics of wild boars and consider the influence of environmental temperature. In connected component analysis, the area threshold is set to 200 pixels to filter out small heat source interference. For the determination of dwelling events, the preset spatial range is defined as a spherical area with a radius of 3 meters, and the preset time threshold is set to 15 seconds. That is, when the animal thermal target stays within the 3-meter range for more than 15 seconds, it is determined as a valid dwelling event, thereby effectively filtering out the animal's brief passing behavior and actual observation behavior.

[0025] Specifically, the acquisition of the position coordinate sequence is achieved through real-time 3D localization using a visual sensor array. Each visual sensor acquires images and extracts the pixel coordinates (x, y, z) of the animal's centroid. i ,y i ), through the preset camera intrinsic parameter matrix [f x ,f y ,c x ,c y The extrinsic parameter matrix [R,T] converts the pixel coordinates into three-dimensional position coordinates (X,Y,Z) in the world coordinate system. A stereo vision algorithm is used to estimate the depth between adjacent visual sensors to eliminate the depth ambiguity of monocular vision. Multiple coordinates at the same time point are fused by weighted average to obtain the final position coordinates, which are sorted by timestamp t to form the position coordinate sequence {(X(t),Y(t),Z(t))|t=0,Δt,2Δt,...}.

[0026] Specifically, the rate of change of movement speed is calculated through the fusion of acoustic and visual localization data. When the acoustic sensor array detects the sound of a target animal, the system calculates the coordinates of the sound source using a hyperbola intersection algorithm through the following steps: First, the timestamp t of the sound signal arriving at each microphone is recorded. i Select a reference microphone as the baseline and calculate the time difference Δt between the other microphones and the reference microphone. ij =t j -t i Then, based on the speed of sound v and the coordinates of each microphone position (x... i ,y i ,z i Establish the system of equations |RR i|-|RR j |=v×Δt ij Where R is the sound source location vector, R i and R j Let X and J be the position vectors of the i-th and j-th microphones, respectively; then, the sound source coordinates (X, J, J) are obtained by solving the nonlinear equations using the least squares method. s ,Y s Z s Finally, the sound source coordinates are matched and verified with the visual position coordinates at the corresponding time, and the displacement change is calculated within a continuous time window to obtain the rate of change of the animal's movement speed.

[0027] Specifically, the residence time distribution is obtained through spatiotemporal statistical analysis of thermal imaging data. The thermal imaging sensor acquires temperature distribution images of the monitored area, and a temperature threshold range of 37℃~40℃ is set to extract the animal's thermal target region. The centroid coordinates of the thermal target region in consecutive frames are calculated to obtain (x... c ,y c A dwell event is recorded when the centroid coordinates fluctuate within a radius of 3 meters for N consecutive frames (N≥150, i.e., 15 seconds). The dwell time is calculated using the formula T. stay =N×0.1 seconds, divide the monitoring area into 10m×10m grid units, and calculate the cumulative dwell time T in each grid unit. total A heatmap of dwell time distribution is generated using an interpolation algorithm, and distribution feature parameters, including the maximum dwell time T, are extracted. max Average stay duration T avg Frequency of stay events F stay By combining spatial trajectory information in the position coordinate sequence and motion state parameters in the rate of change of movement speed, an individual behavior feature dataset {position coordinate sequence, rate of change of movement speed, and residence time distribution} containing spatiotemporal behavioral characteristics is constructed to support subsequent exploratory behavior identification and analysis.

[0028] In one embodiment, such as Figure 2 As shown, in step S1, the animal's exploratory behavior judgment result is generated based on the number of times the animal moves back and forth in the monitoring boundary area, the duration of each stay, and the ratio of entry depth to exit frequency. Set the coordinates of the monitoring boundary line, count the entry and exit events of individual animals crossing the monitoring boundary line in the position coordinate sequence, and accumulate the number of back-and-forth movements; Based on the residence time distribution, the residence events of individual animals in the monitoring area are determined, the start time and end time of each residence event are extracted, and the time difference is calculated to obtain the duration of a single residence. The entry depth is calculated by using the position coordinate sequence to determine the farthest distance from the monitoring boundary line after an animal enters the monitoring area. The exit frequency within a preset time period is calculated by combining the rate of change of movement speed. The ratio of entry depth to exit frequency is then calculated. When the number of back-and-forth movements exceeds the first preset threshold, the duration of a single stay exceeds the second preset threshold, or the ratio of entry depth to exit frequency exceeds the third preset threshold, it is determined to be animal exploratory behavior and output.

[0029] In one embodiment, the coordinates of the monitoring boundary line are set, and the entry and exit events of individual animals crossing the monitoring boundary line in the location coordinate sequence are counted. The cumulative number of back-and-forth movements includes: Based on protection requirements, the spatial coordinate range of the monitoring boundary line within the monitoring area is set, and the monitoring boundary line is represented as a set of continuous line segments; For each coordinate point in the position coordinate sequence, determine its relative positional relationship with the monitoring boundary line. When two consecutive coordinate points are located on opposite sides of the monitoring boundary line, it is marked as a boundary line crossing event. Based on the direction of crossing, boundary line crossing events are classified into entry events and exit events, and the timestamp of each event is recorded. Set a fixed time window length, calculate the absolute value of the difference between the number of entry events and the number of exit events within the window, and when the absolute value of the difference is less than a preset threshold and the total number of entry and exit events is greater than zero, it is considered a valid back-and-forth movement, and the number of back-and-forth movements within different time windows is accumulated.

[0030] Specifically, in this embodiment, the monitoring boundary line is a closed area boundary composed of multiple broken lines, defined by the GPS coordinate point sequence {(X1,Y1),(X2,Y2),...,(X... n ,Y n The diagram shows that each pair of adjacent points forms a boundary line segment. The ray method is used to determine whether the animal's position coordinates are inside the boundary. When performing boundary crossing detection on the position coordinate sequence, the vector cross product method is used to calculate the distance between two consecutive coordinate points P1(X1,Y1,Z1) and P2(X2,Y2,Z2) and the endpoint A(X1,Y1,Z2) of the boundary line segment. a ,Y a ) and B(X b ,Y b The sign of the cross product of the two vectors formed by (P1-A)×(BA) and (P2-A)×(BA) is changed to determine whether a crossing has occurred. When the product of (P1-A)×(BA) and (P2-A)×(BA) is less than zero, it is determined as a crossing event. A fixed time window length of 5 minutes is set. When the absolute value of the difference between the entry event and the exit event within the window is less than 2 and the sum is greater than 4, it is determined as frequent back-and-forth movement behavior. The number of back-and-forth movements is accumulated within 24 hours. When the number exceeds the first preset threshold of 15 times, the first condition for determining the trial behavior is triggered.

[0031] Specifically, in this embodiment, the calculation of a single stay duration is based on stay time distribution data. The system records the start time t of each stay event identified within the monitoring area. start and end time t end Duration of a single stay T stay =t end -t start The unit is seconds; when the spatial location of a dwelling event is no more than 20 meters from the monitoring boundary line, the system marks the dwelling event as a boundary dwelling event; the system calculates the entry depth parameter D. in The vertical distance between the farthest point reached by an individual animal during a complete invasion process (from entry to final exit) and the monitoring boundary line is calculated using the shortest distance between the spatial extreme point in the location coordinate sequence and the boundary line; the exit frequency F out Defined as the number of times an animal exits the monitoring area within a 1-hour observation period, the system calculates the ratio R=D (entry depth to exit frequency). in / F out As a key indicator for measuring exploratory behavior, a second preset threshold is set at 120 seconds. When the duration of a single stay exceeds this threshold, the second condition for judging exploratory behavior is triggered. A third preset threshold is set at 5.0. When the ratio of entry depth to exit frequency exceeds this threshold, the third condition for judging exploratory behavior is triggered. When all three conditions are met simultaneously, the system outputs the animal's exploratory behavior judgment result and triggers an early warning signal.

[0032] In one embodiment, the adaptive parameters of a group to an acoustic and light-induced repulsion stimulus include an initial response threshold, a current response threshold, and a threshold change rate. In step S2, based on the results of the animal's exploratory behavior assessment and historical drive-away event records, the adaptive parameters of the group to the audio-visual drive-away stimuli are calculated using the response threshold reinforcement algorithm, including: Establish a database of historical driving events, recording the time, location, type and intensity of the driving stimulus, and the response behavior of the animal group for each driving event; The initial response threshold is determined based on the intensity of the stimulus when the animal group begins to flee during the first expulsion event; A probabilistic model based on response thresholds was constructed, and the degree of adaptation of animal groups to driving stimuli was analyzed based on the results of animal trial behavior judgment and historical driving event database. Analyze the changing trends of animal groups' responses to the same type of stimuli in historical driving events, calculate the minimum stimulus intensity that causes the group to flee in each driving event, and obtain the current response threshold; The threshold change factor is obtained by calculating the ratio of the current response threshold to the initial response threshold, and the threshold change rate is obtained by calculating the growth rate of the threshold change factor per unit time. Multiply the threshold change factor by a preset reference coefficient, and add the threshold change rate by a preset weight to obtain the population's adaptive parameters to the sound and light driving stimuli.

[0033] It should be noted that the Response Threshold Reinforcement (RTR) algorithm is an adaptive learning algorithm based on the principles of biological group behavior. Its core idea originates from the task allocation mechanism of social insects, applying the changing patterns of individual response thresholds to environmental stimuli to animal herding systems. This algorithm dynamically tracks and quantifies the adaptation process of animal groups to repeated herding stimuli by establishing a mathematical mapping relationship between stimulus intensity and group behavioral response. The algorithm first determines the baseline response threshold based on the initial herding event, then constructs a probability distribution model of the response threshold evolution over time by analyzing stimulus-response pairs in historical herding data. It uses Markov processes to describe changes in group adaptation and employs a Bayesian update method to optimize the model parameters. This invention utilizes this algorithm to not only accurately assess the current group's adaptation to herding stimuli but also predict the decay trend of future herding effects, providing a scientific basis for the dynamic adjustment of herding strategies and effectively solving the adaptation problem faced by traditional fixed-parameter herding methods.

[0034] Specifically, a probabilistic model based on response thresholds is constructed, and the degree of adaptation of the animal population to the driving stimulus is analyzed based on the results of animal trial behavior judgment and historical driving event database, including: Assign an initial response threshold θ to each individual in the animal population. i Establish the individual response probability function P(θ) i ,s), where s represents the intensity of the repulsion stimulus, when s≥θ i Individual i responds to the stimulus; Based on the time, location, and type of driving stimulus records in the historical driving event database, a stimulus-response pairing matrix was constructed to statistically analyze the response behavior patterns of animal groups under different stimulus intensities. Establish a dynamic response threshold update mechanism. When an individual performs escape behavior, the response threshold of that individual is lowered according to the principle of feedback reinforcement, making the individual more sensitive to subsequent stimuli of the same type. The probability distribution of the group response was calculated using the Monte Carlo simulation method, and the results of the animal's exploratory behavior judgment were used as input parameters to adjust the proportion of active individuals in the group. The maximum likelihood estimation method was used to fit the probability density function of the response threshold, and the differences in the sensitivity of the population to different types of driving stimuli were analyzed. Individual response threshold parameters are updated based on Bayesian inference, and the prediction accuracy of the probabilistic model is corrected using newly observed animal group response behavior data. The time evolution trajectory of the average response threshold of the population is calculated to quantify the overall changes in the adaptation of the animal population to the driving stimulus.

[0035] In one embodiment, step S2 involves setting an adaptive critical threshold criterion to generate an animal group drive-off adaptation assessment result, including: Establish criteria for determining adaptive critical thresholds, including upper limits for threshold change multiples and threshold change rates; Compare the threshold change factor with the upper limit of the threshold change factor, and compare the threshold change rate with the upper limit of the threshold change rate; When the threshold change factor exceeds the upper limit of the threshold change factor, or the threshold change rate exceeds the upper limit of the threshold change rate, it is determined that the animal group has adapted to the current driving method, and the adaptive state is marked as adapted; otherwise, it is marked as unadapted. Based on the adaptive status of the markers, the percentage of the group's adaptive level is calculated according to the threshold change factor and threshold change rate, thus obtaining the animal group's drive-off adaptive assessment results.

[0036] Specifically, in this embodiment, the historical expulsion event database adopts a relational data structure design, including fields such as event ID, timestamp, geographic coordinates, expulsion stimulus type (sound / light / odor), stimulus intensity (decibels / luminous flux / concentration), and response behavior (escape / stagnation / ignoring). During the initial expulsion, a gradient-increasing stimulus intensity strategy is used, starting from the lowest stimulus intensity and increasing by one level every 10 seconds until the stimulus intensity value at which more than 70% of the animal group begins to exhibit escape behavior is recorded as the initial response threshold θ0. For sound stimuli, the initial response threshold is set in the range of 85-95 decibels; for light stimuli, the initial response threshold is set in the range of 2000-3000 lumens. By analyzing the changing trend of the minimum stimulus intensity that causes group escape behavior in historical expulsion events, the minimum stimulus intensity required for the nth expulsion event is denoted as θ. n At that time, calculate the ratio κ=θ between the current response threshold and the initial response threshold. n The threshold change factor is obtained by / θ0, and the growth rate of the threshold change factor per unit time μ=dκ / dt is calculated as the threshold change rate. The system multiplies the threshold change factor κ with the preset reference coefficient α (default value 1.5), and adds the threshold change rate μ with the preset weight β (default value 2.0), i.e. S=α×κ+β×μ, to obtain the adaptive parameters of the group to the sound and light driving stimuli.

[0037] Specifically, in this embodiment, the adaptive critical threshold determination criterion sets clear numerical boundaries, including an upper limit κ for the threshold change factor. max Setting it to 2.5 indicates that significant adaptation is considered to have occurred when the animal population's response threshold to a specific stimulus reaches 2.5 times the initial value; the upper limit of the threshold change rate is μ.max The value is set to 0.25 / day, indicating that when the daily growth rate of the animal population's response threshold to a specific stimulus exceeds 25%, it is considered to be in a rapid adaptation state; when the system calculates the real-time adaptation assessment, the degree of adaptation is expressed as a percentage: P=(κ / κ max )×70%+(μ / μ max The system calculates the degree of adaptation by multiplying the result by 30% and then divides it into three levels: low adaptation (P<50%), moderate adaptation (50%≤P<80%), and high adaptation (P≥80%). When P≥80%, the system automatically generates suggestions for adjusting the expulsion strategy, including changing the type of expulsion stimulus, increasing the intensity of the stimulus, or adjusting the frequency pattern of the stimulus.

[0038] To facilitate understanding of the above-mentioned technical solution of the present invention, the following is a specific description using a wild boar invasion prevention and control system at the boundary of a national nature reserve as an example: At the boundary between the protected area and surrounding farmland, frequent wild boar invasions caused crop damage. The local agricultural department decided to implement the intelligent monitoring and deterrence technology of this invention. First, along the boundary between the protected area and the farmland, staff installed a waterproof high-definition network camera every 5 meters, and set up an omnidirectional microphone array (6 microphones in a hexagonal arrangement per array) every 30 meters. Uncooled long-wave infrared thermal imagers were also installed at key monitoring points, and temperature thresholds were set to identify the wild boars as thermal targets.

[0039] During the first month of system operation, local computing equipment processed the data and extracted distribution and behavioral parameters of the wild boar population. Analysis of location coordinate sequences revealed that the wild boar population was primarily active between 10:00 PM and 4:00 AM, frequently moving back and forth at the southeastern boundary of the protected area. The system recorded instances of wild boars crossing the monitoring boundary, showing a cumulative total of 18 such crossings within 24 hours. Simultaneously, thermal imaging data analysis showed that wild boars lingered in the grass 15 meters from the boundary for 180 seconds. Combining this with location data, the system calculated the maximum entry depth to be 25 meters, while the exit frequency was 3 times per hour, resulting in an entry depth to exit frequency ratio of 8.33, exceeding the preset threshold of 5.0. Based on these three indicators, the system generated a determination of the wild boars' exploratory behavior, confirming that the wild boar population was purposefully exploring the farmland.

[0040] Subsequently, the system initiated the expulsion control process. During the initial expulsion, the system started with the lowest stimulus intensity, using the call of a wild boar's natural predator (wolf) at 85 decibels as the sound stimulus. It was observed that over 70% of the wild boars began to flee; this intensity was recorded as the initial response threshold. After two weeks of continuous expulsion, the system found that the minimum sound stimulus intensity required to cause the wild boar herd to flee had increased to 105 decibels, with an adaptation assessment P=36.76%, determining that the current wild boar herd was in a state of "low adaptation."

[0041] Based on the above-mentioned exploratory behavior judgment results and adaptation assessment results, the system generated corresponding driving control signals. The control signals mainly consist of two parts: a start command and adjustment parameters. The start command determined the driving device numbers to be activated (six devices E15-E20 at the southeast corner of the reserve), the device activation order (starting from the device closest to the wild boar group, E17, and expanding outwards to E15, E16, E18, E19, and E20), and the scheduled execution time (22:30, i.e., before the peak period of wild boar activity). The adjustment parameters included sound stimulus parameters (selecting wolf howls as the sound type, with a sound pressure level set to 95 dB, higher than the initial response threshold but lower than the current response threshold) and light stimulus parameters (white light, brightness 2000 lumens, flicker frequency 3 Hz). Since the adaptation assessment showed that the wild boar group was in a "low-adaptation" state, the system selected a combined stimulus mode rather than a single sound stimulus to avoid rapid adaptation of the wild boar group to a single stimulus type. The driving control signal is sent in encrypted data packets to the controllers of each driving device on site via the system's communication module. After receiving the signal, the controller executes the driving operation according to the specified parameters. This adaptive control strategy, based on animal behavior characteristic analysis and adaptive assessment, enables the system to dynamically adjust the driving method according to changes in the behavior of the wild boar group. This effectively avoids the adaptation problems that are easily caused by traditional fixed-mode driving, and significantly improves the effectiveness and durability of animal invasion prevention and control.

[0042] After three months of system operation, wild boar invasions in the southeast corner of the reserve decreased by 87%, and crop losses decreased by 92%. This successful case demonstrates that the present invention, by combining animal behavioral characteristic analysis, adaptive assessment, and adaptive control technology, achieves intelligent, efficient, and sustainable prevention and control of wild animal invasions.

[0043] like Figure 3 As shown, according to another embodiment of the present invention, a multi-source array intelligent animal monitoring and deterrence system based on target recognition is also provided. This multi-source array intelligent animal monitoring and deterrence system based on target recognition includes: Individual exploration module 1 is used to extract the distribution behavior parameters of individual animals based on the data collected by the multi-source sensor array, obtain the individual behavior feature dataset, and generate the animal exploration behavior judgment result based on the number of times the individual animal moves back and forth in the monitoring boundary area, the duration of a single stay, and the ratio of entry depth to exit frequency. Group adaptation module 2 is used to calculate the group's adaptation parameters to sound and light driving stimuli based on the results of animal trial behavior judgment and historical driving event records, using the response threshold enhancement algorithm, and to set the adaptation critical threshold judgment criteria to generate animal group driving adaptation assessment results. The regulation and control module 3 is used to generate a deportation control signal containing a start command and regulation parameters based on the animal's exploratory behavior judgment results and the animal group's deportation adaptability assessment results.

[0044] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-source array-based intelligent animal monitoring and deterrence method based on target recognition, characterized in that, include: S1. Based on the data collected by the multi-source sensor array, extract the distribution behavior parameters of individual animals to obtain an individual behavior feature dataset, and generate the animal exploration behavior judgment result according to the number of times the individual animal moves back and forth in the monitoring boundary area, the duration of a single stay, and the ratio of entry depth to exit frequency. S2. Based on the results of animal trial behavior and historical driving event records, the adaptive parameters of the group to the sound and light driving stimuli are calculated using the response threshold enhancement algorithm, and the adaptive critical threshold judgment criteria are set to generate the animal group driving adaptation assessment results. S3. Based on the results of animal trial behavior assessment and animal group drive-off adaptability assessment, generate a drive-off control signal containing activation instructions and adjustment parameters.

2. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 1, characterized in that, The collected data includes visual, acoustic, and thermal imaging data; the distribution behavior parameters include position coordinate sequences, rate of change of movement speed, and residence time distribution. The data collected based on the multi-source sensor array is used to extract the distribution and behavioral parameters of individual animals, resulting in an individual behavioral feature dataset including: The pixel coordinates of individual animals in the image are obtained through visual data. The pixel coordinates are converted into visual position coordinates according to the preset camera calibration matrix, and the spatial coordinate changes are recorded in time series to form a position coordinate sequence. Based on acoustic data, the sound of the target animal is identified, and the time difference of the sound reaching each acoustic sensor is calculated to determine the location of the sound source. The sound source location is then correlated and matched with the location coordinate sequence to solve for the rate of change of movement speed. Animal thermal targets are marked based on thermal imaging data. The duration of the centroid coordinates of the animal thermal targets is counted to determine dwell events. The dwell time distribution of animal individuals in the monitoring area is calculated. Combined with the position coordinate sequence and the rate of change of movement speed, an individual behavior feature dataset is established.

3. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 2, characterized in that, The process of acquiring the pixel coordinates of an individual animal in an image through visual data, converting the pixel coordinates into visual position coordinates according to a preset camera calibration matrix, and recording the spatial coordinate changes in a time series to form a position coordinate sequence includes: Based on image data collected by different visual sensors within the monitoring area, inter-frame difference is performed on each visual sensor image to extract the moving target region; The extracted moving target region is matched and identified using animal body feature templates, the individual animal contour region is marked, and the pixel coordinates of the contour centroid are extracted. The pixel coordinates extracted from each visual sensor are converted into visual position coordinates in a unified three-dimensional coordinate system through a pre-calibrated camera intrinsic and extrinsic parameter matrix. Spatial coordinates obtained from different visual sensors at the same time point are weighted and fused to eliminate viewpoint errors and form a time-stamped sequence of location coordinates.

4. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 2, characterized in that, The process of identifying target animal sounds based on acoustic data, calculating the time difference between the arrival of the animal's sounds at each acoustic sensor to determine the sound source location, associating and matching the sound source location with the location coordinate sequence, and solving for the rate of change of movement speed includes: Based on an acoustic sensor array, sound signals generated by animal activities are collected and preprocessed to eliminate background noise; Extract the time-frequency features of the sound signal and identify the target animal's sound based on a pre-set animal sound feature database; The time difference between the arrival of the target animal's sound at each acoustic sensor is calculated, and the spatial coordinates of the sound source are calculated using a sound source localization algorithm, taking into account the spatial distribution parameters of the acoustic sensors. The spatial coordinates of the sound source are associated and matched with the visual position coordinates at the corresponding time points. The displacement distance between adjacent sampling time points is calculated by the position coordinate sequence within the time window and divided by the time interval to obtain the rate of change of the animal's movement speed in each time period.

5. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 2, characterized in that, The process of marking individual animal thermal targets based on thermal imaging data, statistically analyzing the duration of the centroid coordinates of individual animal thermal targets to determine dwell events, and calculating the dwell time distribution of individual animals within the monitoring area includes: Based on the collected thermal imaging data of the monitoring area, upper and lower limits of temperature thresholds are set according to the body temperature characteristics of the target animal. The thermal imaging data is binarized and thermal signal areas that meet the temperature threshold range are extracted. Connectivity analysis is performed on the thermal signal region to remove noise regions smaller than a preset area threshold, and thermal signal regions that meet the size condition are marked as thermal targets of individual animals. Calculate the centroid coordinates of the thermal target of an individual animal and establish a mapping relationship between the centroid coordinates and the position coordinate sequence; Based on the mapping relationship, the duration of the thermal target of an individual animal in a continuous frame sequence is statistically analyzed, and the corresponding three-dimensional coordinates are recorded. When the centroid coordinates are within a preset spatial range and the duration exceeds a preset time threshold, it is determined as a dwell event. The total duration of dwell time for each three-dimensional coordinate is accumulated to form a heat map of the dwell time distribution of individual animals within the monitoring area, and the dwell time distribution is extracted.

6. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 1, characterized in that, The animal exploration behavior judgment result is generated based on the number of times an individual animal moves back and forth in the monitoring boundary area, the duration of each stay, and the ratio of entry depth to exit frequency. Set the coordinates of the monitoring boundary line, count the entry and exit events of individual animals crossing the monitoring boundary line in the position coordinate sequence, and accumulate the number of back-and-forth movements; Based on the residence time distribution, the residence events of individual animals in the monitoring area are determined, the start time and end time of each residence event are extracted, and the time difference is calculated to obtain the duration of a single residence. The entry depth is calculated by using the position coordinate sequence to determine the farthest distance from the monitoring boundary line after an animal enters the monitoring area. The exit frequency within a preset time period is calculated by combining the rate of change of movement speed. The ratio of entry depth to exit frequency is then calculated. When the number of back-and-forth movements exceeds the first preset threshold, the duration of a single stay exceeds the second preset threshold, or the ratio of entry depth to exit frequency exceeds the third preset threshold, it is determined to be animal exploratory behavior and output.

7. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 6, characterized in that, The process of setting the coordinates of the monitoring boundary line, counting the entry and exit events of individual animals crossing the monitoring boundary line in the position coordinate sequence, and accumulating the number of back-and-forth movements includes: Based on protection requirements, the spatial coordinate range of the monitoring boundary line within the monitoring area is set, and the monitoring boundary line is represented as a set of continuous line segments; For each coordinate point in the position coordinate sequence, determine its relative positional relationship with the monitoring boundary line. When two consecutive coordinate points are located on opposite sides of the monitoring boundary line, it is marked as a boundary line crossing event. Based on the direction of crossing, boundary line crossing events are classified into entry events and exit events, and the timestamp of each event is recorded. Set a fixed time window length, calculate the absolute value of the difference between the number of entry events and the number of exit events within the window, and when the absolute value of the difference is less than a preset threshold and the total number of entry and exit events is greater than zero, it is considered a valid back-and-forth movement, and the number of back-and-forth movements within different time windows is accumulated.

8. The multi-source array animal intelligent monitoring and repelling method based on target recognition according to claim 1, characterized in that, The adaptive parameters of the population to the sound and light driving stimuli include the initial response threshold, the current response threshold, and the threshold change rate. The adaptive parameters of the group to the sound and light stimuli, calculated using the response threshold reinforcement algorithm based on the results of animal trial behavior assessment and historical drive-away event records, include: Establish a database of historical driving events, recording the time, location, type and intensity of the driving stimulus, and the response behavior of the animal group for each driving event; The initial response threshold is determined based on the intensity of the stimulus when the animal group begins to flee during the first expulsion event; A probabilistic model based on response thresholds was constructed, and the degree of adaptation of animal groups to driving stimuli was analyzed based on the results of animal trial behavior judgment and historical driving event database. Analyze the changing trends of animal groups' responses to the same type of stimuli in historical driving events, calculate the minimum stimulus intensity that causes the group to flee in each driving event, and obtain the current response threshold; The threshold change factor is obtained by calculating the ratio of the current response threshold to the initial response threshold, and the threshold change rate is obtained by calculating the growth rate of the threshold change factor per unit time. Multiply the threshold change factor by a preset reference coefficient, and add the threshold change rate by a preset weight to obtain the population's adaptive parameters to the sound and light driving stimuli.

9. A multi-source array-based intelligent animal monitoring and deterrence method based on target recognition according to claim 8, characterized in that, The established adaptive threshold criteria, used to generate animal group drive-off adaptation assessment results, include: Establish criteria for determining adaptive critical thresholds, including upper limits for threshold change multiples and threshold change rates; Compare the threshold change factor with the upper limit of the threshold change factor, and compare the threshold change rate with the upper limit of the threshold change rate; When the threshold change factor exceeds the upper limit of the threshold change factor, or the threshold change rate exceeds the upper limit of the threshold change rate, it is determined that the animal group has adapted to the current driving method, and the adaptive state is marked as adapted; otherwise, it is marked as unadapted. Based on the adaptive status of the markers, the percentage of the group's adaptive level is calculated according to the threshold change factor and threshold change rate, thus obtaining the animal group's drive-off adaptive assessment results.

10. A multi-source array animal intelligent monitoring and deterrence system based on target recognition, used to implement the multi-source array animal intelligent monitoring and deterrence method based on target recognition as described in any one of claims 1-9, characterized in that, The system includes: The individual exploration module is used to extract the distribution behavior parameters of individual animals based on the data collected by the multi-source sensor array, obtain the individual behavior feature dataset, and generate the animal exploration behavior judgment result based on the number of times the individual animal moves back and forth in the monitoring boundary area, the duration of a single stay, and the ratio of entry depth to exit frequency. The group adaptation module is used to calculate the group's adaptation parameters to the sound and light driving stimuli based on the results of animal trial behavior judgment and historical driving event records, using the response threshold enhancement algorithm, and to set the adaptation critical threshold judgment criteria to generate animal group driving adaptation assessment results. The regulation and control module is used to generate a deportation control signal containing activation commands and regulation parameters based on the results of animal exploratory behavior judgment and animal group deportation adaptability assessment.

Images