Field panda behavior monitoring method and system based on artificial intelligence

Abstract

The invention provides a field panda behavior monitoring method and system based on artificial intelligence, and relates to the technical field of animal behavior monitoring, and the method comprises the steps: obtaining an original data set of a target panda; performing feature extraction processing according to the video information and the environment information, and analyzing feature change trends in different time periods and different scenes by extracting motion, posture and micro expression information of a target panda and combining environment conditions to obtain a feature set; carrying out numeralization processing according to the feature set to obtain virtual individual numerical value representation; modeling processing is carried out according to the virtual individual numerical value representation, and a panda motor function model is constructed; performing behavior trajectory deduction according to the panda motor function model and the geographic position information to obtain a behavior trajectory deduction result; and generating a behavior monitoring report of the target panda based on the behavior track deduction result. According to the method, the behavior pattern of the panda can be analyzed more comprehensively and accurately, and the monitoring efficiency and the protection effect are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of animal behavior monitoring, and in particular to a method and system for monitoring the behavior of wild giant pandas based on artificial intelligence. Background Art

[0002] Pandas are one of the rarest and most endangered wild animals in the world and are flagship species in conservation biology. With the increasing demand for wildlife conservation and ecological research, wild giant panda behavior monitoring technology has gradually become one of the key technologies for biological research, ecological protection and behavioral analysis. Existing giant panda behavior monitoring methods mainly rely on manual observation and video monitoring, with observers recording the activity trajectories and behavioral characteristics of animals. However, these methods have many problems: first, manual observation is inefficient and easily affected by the subjective factors of the observer; second, although video monitoring can provide data, its processing relies on manual analysis, and it is often unable to effectively capture and predict the dynamic behavior patterns of giant pandas in real time. In addition, the existing technologies have deficiencies in the processing of data across monitoring points, especially under the influence of time asynchrony and environmental interference, the data collected from multiple monitoring points lack effective synchronization, resulting in inaccurate and inconsistent analysis results. These existing technologies cannot comprehensively and systematically analyze complex factors such as the spatiotemporal changes and emotional fluctuations of giant panda behavior, and therefore have limitations in behavior prediction and dynamic monitoring. Existing technologies have not been able to effectively combine the animals' movement, health and emotional states into a unified model, and they mostly rely on manual analysis and lack automated and accurate behavior prediction capabilities, resulting in the inability to accurately reflect the giant pandas' dynamic behavior and health status in real time.

[0003] Based on the above shortcomings of the prior art, there is an urgent need for a wild giant panda behavior monitoring method and system based on artificial intelligence. Summary of the invention

[0004] The purpose of the present invention is to provide a method for monitoring the behavior of wild giant pandas based on artificial intelligence to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows: In the first aspect, the present application provides a method for monitoring the behavior of wild giant pandas based on artificial intelligence, comprising: Acquire an original data set of a target giant panda, wherein the original data set is video information, geographic location information, and environmental information of the same wild giant panda collected from multiple monitoring locations; Performing feature extraction processing according to the video information and the environmental information, extracting the action, posture and micro-expression information of the target giant panda, and analyzing the feature change trends in different time periods and different scenes in combination with the environmental conditions, to obtain a feature set; Numerical processing is performed according to the feature set, and the behavior characteristics, health characteristics and emotional characteristics are quantified into activity intensity index, body balance index and emotional fluctuation index respectively to obtain a virtual individual numerical representation; Modeling is performed according to the virtual individual numerical representation, and a giant panda movement function model is constructed by simulating the skeletal movement posture and combining the influence of health and emotional state on individual behavior selection; Deducing the behavior trajectory according to the giant panda's motor function model and the geographic location information, simulating the dynamic behavior selection process of the giant panda in different scenarios using a time series model and a Markov decision process, and obtaining the behavior trajectory deduction result; A behavior monitoring report of the target giant panda is generated based on the behavior trajectory deduction results.

[0005] In the second aspect, the present application also provides a wild giant panda behavior monitoring system based on artificial intelligence, including: An acquisition module is used to acquire an original data set of a target giant panda, wherein the original data set is video information, geographic location information, and environmental information of the same wild giant panda collected from multiple monitoring locations; An extraction module is used to perform feature extraction processing according to the video information and the environmental information, extract the action, posture and micro-expression information of the target giant panda, and analyze the feature change trends in different time periods and different scenes in combination with the environmental conditions to obtain a feature set; A conversion module, used for performing numerical processing according to the feature set, and obtaining a virtual individual numerical representation by quantifying the behavior characteristics, health characteristics and emotional characteristics into an activity intensity index, a body balance index and an emotional fluctuation index respectively; A modeling module is used to perform modeling processing according to the virtual individual numerical representation, and construct a giant panda movement function model by simulating the skeletal movement posture and combining the influence of health and emotional state on individual behavior selection; A deduction module is used to deduce the behavior trajectory according to the giant panda's motor function model and the geographic location information, and to simulate the dynamic behavior selection process of the giant panda in different scenarios by using a time series model and a Markov decision process to obtain a behavior trajectory deduction result; The output module generates a behavior monitoring report of the target giant panda based on the behavior trajectory deduction results.

[0006] The beneficial effects of the present invention are: The present invention obtains multi-source data, including surveillance videos, geographic location and environmental information, and extracts features from the data to obtain the behavioral, health and emotional feature sets of giant pandas. Through virtual individual modeling, the behavioral characteristics and emotional fluctuations of giant pandas are digitized, motor function models and decision-making models are established, and behavioral trajectory deduction is achieved. This enables a more comprehensive and accurate analysis of the giant panda's behavioral patterns, improves the efficiency of monitoring and protection of giant pandas, and effectively solves the problems of data synchronization across monitoring points and environmental interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0008] Figure 1 A flowchart of a method for monitoring wild giant panda behavior based on artificial intelligence described in an embodiment of the present invention; Figure 2 Schematic diagram of the structure of the wild giant panda behavior monitoring system based on artificial intelligence described in an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of the wild giant panda behavior monitoring device based on artificial intelligence described in an embodiment of the present invention.

[0009] Markings in the figure: 800, wild giant panda behavior monitoring equipment based on artificial intelligence; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component; 901, acquisition module; 902, extraction module; 903, conversion module; 904, modeling module; 905, deduction module; 906, output module. DETAILED DESCRIPTION

[0010] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0011] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0012] Embodiment 1:

[0013] This embodiment provides a method for monitoring the behavior of wild giant pandas based on artificial intelligence.

[0014] See also Figure 1 , the figure shows that the method includes steps S100 to S500.

[0015] Step S100, obtaining an original data set of a target giant panda, where the original data set is video information, geographic location information, and environmental information of the same wild giant panda collected from multiple monitoring locations; It needs to be explained that the video information provides dynamic behavior data of giant pandas at different time points, including key characteristics such as their activity range, behavior patterns, and posture changes; the geographic location information accurately records the spatial distribution and activity trajectory of giant pandas in the ecological environment through GPS or other positioning technologies; environmental information includes environmental variables such as temperature, humidity, air pressure, and vegetation density.

[0016] Step S200, performing feature extraction processing according to the video information and environmental information, extracting the action, posture and micro-expression information of the target giant panda, and analyzing the feature change trends in different time periods and different scenes in combination with environmental conditions to obtain a feature set; It is understandable that action feature extraction analyzes the dynamic behavior of giant pandas in the video, usually including their movement status, behavior type (such as foraging, resting, moving, socializing, etc.) and activity intensity. Body feature extraction focuses on the posture and morphological changes of giant pandas, especially the transitions of their walking, standing, crouching and other postures. Micro-expression feature extraction focuses on the subtle changes in the facial expressions of giant pandas, especially through facial recognition technology, capturing details such as slight movements of the ears and changes in the eyes. These subtle facial changes can reflect the emotional state of giant pandas, such as anxiety, relaxation or pleasure, and thus affect their subsequent behavioral choices. Combined with environmental factors (such as temperature, humidity or food resources), these emotional features provide a deep emotional fluctuation background for analyzing the behavior of giant pandas.

[0017] Step S300, performing numerical processing according to the feature set, by quantifying the behavior characteristics, health characteristics and emotional characteristics into activity intensity index, body balance index and emotional fluctuation index respectively, to obtain a virtual individual numerical representation; Specifically, first, the quantification of behavioral characteristics is to convert the intensity, frequency and duration of giant pandas' activities into an activity intensity index. This quantification process needs to be combined with motion data in video analysis, including the giant panda's movement speed, activity range and movement pattern in a specific time period. The quantification of health characteristics focuses on the giant panda's physical condition, especially its body balance. The body balance index is calculated by extracting the skeleton key points, posture changes and body contact with the ground in the video. This index can not only quantify the giant panda's physical health, but also reflect its adaptation to balance and stability in different environments. The quantification of emotional characteristics obtains the emotional fluctuation index through the analysis of facial expressions and behavioral changes. For example, the facial expression analysis model extracts the micro-expression changes of giant pandas, and combines the influence of environmental factors on emotions to obtain the emotional fluctuation index. This index can reflect the giant panda's psychological state, such as anxiety, relaxation or joyful emotional fluctuations, which further affects its behavioral choices and activity patterns.

[0018] Step S400, modeling is performed according to the virtual individual numerical representation, by simulating the skeleton movement posture and combining the influence of health and emotional state on individual behavior selection, to construct a giant panda movement function model; It should be noted that this step can not only accurately capture the movement trajectory of giant pandas, but also deeply reveal the physiological and emotional factors behind their behavioral choices by integrating skeletal movement simulation and health and emotional state analysis. This enables the model to more realistically reflect the behavior of giant pandas in complex environments, providing a basis for subsequent behavior trajectory deduction and prediction.

[0019] Step S500, performing behavior trajectory deduction based on the giant panda's motor function model and geographic location information, using a time series model and a Markov decision process to simulate the giant panda's dynamic behavior selection process in different scenarios, and obtaining a behavior trajectory deduction result; It can be understood that, firstly, this step uses the time series model to capture the changes in the behavior patterns of giant pandas in different time periods and environmental conditions. Through the training of the time series model, it is possible to predict its behavior trends in a specific time period in the future based on historical data, so that the behavior trajectory deduction has a stronger time series correlation. Secondly, the Markov decision process is used to simulate the decision-making process of giant pandas in complex environments. Specifically, the Markov model simulates how giant pandas make optimal behavioral choices based on the current state by defining the relationship between state, action, and reward. Based on this model, the probability of giant pandas making transitions between different behaviors can be described, and their behavioral paths can be further simulated.

[0020] Step S600: Generate a behavior monitoring report of the target giant panda based on the behavior trajectory deduction result.

[0021] It should be noted that through in-depth analysis of the giant panda's behavioral trajectory, the report can provide the giant panda's activity patterns, behavioral change trends, and behavioral adaptability in different environments within a specific time period, providing scientific and systematic behavioral prediction information for nature reserve managers or researchers.

[0022] Further, step S200 includes step S210 to step S240.

[0023] Step S210, performing motion feature extraction processing according to the video information, performing sparse decomposition on the motion features in the video by using a sparse coding algorithm, extracting key frames and motion patterns of the giant panda's motion trajectory to construct a motion feature dictionary, and obtaining a motion feature set; It can be understood that the application of the sparse coding algorithm is to represent the action features of the giant panda in the video as a set of sparse basis vectors. These basis vectors can capture the key information of the giant panda in different behavior modes, such as walking, standing, crawling, eating, etc. The advantage of sparse coding is that it can effectively reduce the redundancy of data, represent complex action sequences as a small number of basis vectors and corresponding weights, and thus extract the key features of the giant panda's behavior. Sparse coding can not only extract representative actions in the giant panda's motion trajectory, but also identify the most representative key frames in the motion process. These key frames reflect the important turning points of the giant panda in the motion process, such as the transition from stillness to motion, or the switching of motion methods. The final constructed action feature dictionary can not only represent the basic actions of the giant panda, but also reflect its behavioral changes in different environments, and has a high degree of generalization.

[0024] Step S220: performing posture feature extraction processing according to the video information, performing dimension reduction and feature decomposition on the posture information in the video by applying a non-negative matrix decomposition algorithm, extracting its posture changes in different scenes, and obtaining a posture feature set; It is understandable that the non-negative matrix factorization algorithm can decompose high-dimensional body posture information into several meaningful, non-negative basis vectors, and represent the body posture changes of giant pandas in different scenes by combining these basis vectors. Body posture changes include actions such as standing, walking, crawling, sitting and lying, which directly affect the behavior and health status of giant pandas. Through the non-negative matrix factorization algorithm, the body posture characteristics of giant pandas can be efficiently extracted from video data, reducing noise and redundant information, thereby improving the accuracy of analysis.

[0025] Step S230, performing micro-expression feature extraction processing based on the video information, extracting local features of ear shaking and eye expression changes in the facial area of ​​the giant panda through a local binary pattern algorithm, analyzing the emotional state in combination with a preset typical facial structure and expression pattern, and obtaining a micro-expression feature set; Specifically, the local binary pattern algorithm is used to analyze tiny expressions such as ear shaking and eye changes in the panda's facial area. These subtle changes reflect the panda's emotional state. By extracting local features of the facial area, the local binary pattern algorithm can capture subtle patterns of change, such as micro-expressions of anxiety, joy, tension and other emotions. These micro-expression changes are not easy to detect, but have a profound impact on the panda's behavioral choices. For example, when a panda is in an anxious state, it may choose to leave certain areas, while when it is in a happy mood, it may stay in a certain place for a longer time.

[0026] Step S240, perform feature integration processing based on the action feature set, posture feature set and micro-expression feature set, analyze the impact of environmental factors on the behavior of giant pandas by combining the action, posture and micro-expression features with environmental information, and obtain a comprehensive feature set through feature weighting and normalization processing.

[0027] It should be noted that by weighting and normalizing these features, we ensure that the contribution of each feature in data integration is reasonably balanced to avoid the bias of a certain feature on the model. The final comprehensive feature set can fully reflect the behavior, health and emotional state of giant pandas.

[0028] Further, step S300 includes step S310 to step S340.

[0029] Step S310: Perform behavior timing alignment processing according to the action features in the feature set, perform nonlinear alignment on action sequences of unequal lengths across monitoring points through a dynamic time warping algorithm, and introduce ambient light intensity as a time expansion penalty factor to calculate the activity intensity index; It is understandable that the purpose of behavioral timing alignment processing is to solve the problem of temporal asynchrony of data across monitoring points, especially when the action sequences collected at different monitoring locations have inconsistent lengths and asynchronous timestamps. To achieve this goal, a dynamic time warping algorithm is used to nonlinearly align action sequences of unequal lengths across monitoring points. The dynamic time warping algorithm can deform and align action sequences on the time axis so that similar action patterns in the time sequence can be aligned. In order to further improve the accuracy of timing alignment, the ambient light intensity is introduced as a time expansion penalty factor, so that the time axis can be reasonably adjusted according to the changes in light, taking into account that the light intensity may affect the execution rate of giant panda behavior (for example, in a low-light environment, the giant panda's activity speed may be slower). After this processing, the activity intensity index obtained can effectively reflect the activity level of giant pandas under different environmental conditions.

[0030] Step S320: Performing posture space embedding processing according to the posture features in the feature set, using an isometric feature mapping algorithm to map the skeleton key points extracted from the multi-view video to a low-dimensional manifold space, and obtaining a posture balance index by calculating the covariance matrix of the principal curvature change rate and the ground contact pressure; It should be noted that the posture space embedding processing focuses on how to extract accurate posture balance information from the key point data of the giant panda skeleton extracted from multi-view videos. By adopting the isometric feature mapping algorithm, the high-dimensional skeleton data is mapped to the low-dimensional manifold space, thereby simplifying the representation of the posture data and maintaining important posture information. This method can capture the movement balance and stability of the giant panda by calculating the relative position changes of each key point of the skeleton. Furthermore, by calculating the covariance matrix of the change rate of the skeleton's principal curvature and the ground contact pressure, the balance of the giant panda's posture can be quantitatively evaluated to generate a posture balance index. This index can reflect the coordination of the giant panda during movement, especially its movement performance in complex terrain or dynamic environments. The calculation formula for the posture balance index is as follows: ; Among them, B is the body balance index, which measures the balance of giant pandas under different environmental and behavioral conditions; t is the sequence number of the time; ∂ is the partial derivative; k(t) is the time function of the principal curvature, which represents the change of the giant panda's skeleton curvature at time t; P(t) is the time function of the ground contact pressure, which represents the contact pressure between the skeleton and the ground at time t; κ i is the principal curvature of the i-th sample point, indicating the change of the skeleton posture; P i is the ground contact pressure of the i-th sample point; n is the total number of sample points; t0 and t1 are time intervals, which are used to consider the dynamic effect of body posture changes over a period of time; v is the weight parameter; λ is the regularization parameter; A and B are the posture change matrix and contact pressure matrix of the giant panda skeleton, respectively, ∥A∥ F and ∥B∥ F is the Frobenius norm of A and B, indicating the overall change intensity; μ is the regularization parameter; C is the covariance matrix, which is used to represent the interaction between the principal curvature and the contact pressure; det(C) is the determinant of the C matrix.

[0031] Step S330, performing emotion coupling analysis and processing according to the micro-expression features in the feature set, applying fuzzy cognitive graph to model the nonlinear relationship between micro-expression electromyographic signals, environmental temperature and humidity, and eating behavior, determining the phase change threshold of the emotional state by calculating the Lyapunov exponent spectrum of the attractor domain, and obtaining the emotion fluctuation index; Specifically, the use of fuzzy cognitive graph modeling can model the relationship between the electromyographic signals of micro-expressions and environmental changes (such as temperature and humidity) and the behavior of giant pandas (such as eating behavior). By introducing the analysis framework of fuzzy cognitive graphs, complex nonlinear connections can be established between emotions, environment and behavior, thereby more accurately quantifying the impact of emotional fluctuations on the behavioral choices of giant pandas. By calculating the Lyapunov index spectrum of the attractor domain, the phase transition threshold of the emotional state can be determined, the critical point of the giant panda's emotional change can be predicted, and then the emotional fluctuation index can be obtained.

[0032] Step S340: Perform multimodal fusion processing according to the activity intensity index, body balance index and mood fluctuation index, perform Pareto front search on all indexes, and generate a virtual individual numerical representation by defining a fitness function including an energy consumption function and a social contact probability.

[0033] It should be noted that Pareto frontier search can optimize multiple objectives to ensure that the optimal balance point is found between different indexes, avoiding any one feature from having an excessive impact on the final result. At the same time, the defined fitness function includes energy consumption function and social contact probability, which helps to optimize the behavior model of the virtual individual so that it can minimize energy consumption while taking into account the interaction frequency of giant pandas in social behavior. The fitness function is expressed as: ; Among them, F is the fitness function, which represents the objective function of giant panda behavior optimization; d represents differential; I activity (t) is the activity intensity index, which indicates the activity intensity of the giant panda at time t; I emotion (t) is the mood fluctuation index, which indicates the intensity of the giant panda’s mood fluctuation at time t; S contact (t) is the social contact probability, which indicates the contact frequency between giant pandas and other individuals; E consumption (t) is the energy consumption function, which represents the energy consumption of the giant panda at time t; α, β, γ and δ are all weight coefficients; σ is the time decay factor, which is used to control the changing trend of energy consumption over time; T represents the time range; e is the base of the natural logarithm.

[0034] Further, step S400 includes step S410 to step S440.

[0035] Step S410, performing skeletal dynamics modeling processing according to the virtual individual numerical representation, using differential algebraic equations to describe joint motion constraints, and obtaining a rigid-flexible coupling skeletal model; It is understandable that differential algebraic equations can accurately simulate the dynamic behavior of the giant panda skeleton by considering factors such as the constraints between joints, range of motion, and moment of inertia. Differential algebraic equations can not only describe the relative motion of the various skeletal parts of the giant panda, but also capture the complex interactions between bones in actual movements, such as moving, standing, walking, jumping, etc. By constructing a rigid-flexible coupled skeleton model, the rigidity (stiffness of bones) and flexibility (flexibility of joints and soft tissues) of the skeleton can be considered at the same time, making the simulation results more in line with the real situation.

[0036] Step S420: Perform health status integration processing according to the rigid-flexible coupling skeleton model, apply the stochastic optimal control theory to establish a metabolic energy-exercise efficiency conversion model, describe the probability density evolution of health indicators through the Focke-Planck equation, and obtain a set of physiological constraint exercise strategies; It should be noted that the stochastic optimal control theory aims to describe the conversion relationship between energy consumption and exercise efficiency during the giant panda's exercise. Through the stochastic optimal control theory, the model can take into account the energy optimization process of the giant panda when performing exercise tasks, that is, completing the necessary behavior without consuming too much energy. At the same time, the Fokker-Planck equation is used to describe the probability density evolution of health indicators, taking into account the physiological changes of giant pandas in different health states, which in turn affects their exercise efficiency and behavioral choices.

[0037] Step S430, performing emotion-behavior coupling processing according to the physiological constraint movement strategy set, and obtaining an emotion enhancement decision tree by modeling the influence intensity of emotion fluctuations on path selection; It can be understood that this step can predict how emotional changes affect the behavioral decisions of giant pandas in different environments by modeling the intensity of the impact of emotional fluctuations on the path selection of giant pandas. For example, when a giant panda is in an anxious state, its behavioral path may tend to a safe area, while when it is relaxed or happy, its movement path may be wider and more exploratory. This process models the relationship between emotional fluctuations and behavioral choices by establishing an emotional enhancement decision tree. The emotional enhancement decision tree can generate a possible behavioral selection tree for giant pandas under the influence of their emotional fluctuations, thereby taking into account the influence of emotional state in the simulation and providing a more detailed decision-making basis for behavioral deduction.

[0038] Step S440: Perform multi-scale fusion processing according to the emotion enhancement decision tree to obtain a giant panda motor function model.

[0039] The core of this step is to integrate decision-making information at different levels, including behavioral patterns, health status, and emotional fluctuations, to form a multi-dimensional movement decision-making framework, which fully reflects the behavior patterns of target giant pandas in different environments and emotional states, and provides accurate model support for subsequent behavior deduction and monitoring.

[0040] Further, step S500 includes step S510 to step S540.

[0041] Step S510: Perform terrain manifold modeling according to the geographical location information, use a diffusion mapping algorithm to fuse the slope, bamboo density and stream distribution data, calculate the terrain similarity matrix and combine the joint torque constraints to generate a terrain accessibility map; In this step, the scatter mapping algorithm not only considers the complexity of the terrain in which the giant panda is located, but also evaluates the accessibility of different terrain areas by calculating the terrain similarity matrix. In addition, combined with joint torque constraints, the algorithm can take into account the movement limitations of its bones and muscles when simulating the giant panda's behavior, thereby more realistically reflecting the giant panda's range of activities. The resulting terrain accessibility map provides an environment-based feasibility framework for the giant panda's behavior deduction, which can accurately represent the potential paths of the giant panda's movement in complex terrain.

[0042] Step S520: Discretize the spatiotemporal state according to the terrain accessibility map, apply the non-uniform Poisson process to establish a behavior event flow model, and use the Hawkes process to model the self-motivation effect of food discovery events on the movement path to obtain a potential behavior transfer network; Specifically, first, a behavioral event flow model is established using the non-uniform Poisson process, which can capture the flow characteristics of giant pandas' behavioral events under different spatiotemporal states. The non-uniform Poisson process can flexibly model the occurrence of events at irregular time intervals, and is particularly suitable for describing the behavioral dynamics of giant pandas in complex environments. In addition, the use of the Hawkes process to model the self-motivational effect of food discovery events on the movement path of giant pandas can accurately describe how giant pandas' behaviors change rapidly after discovering food and stimulate a series of subsequent behavioral choices. Through the modeling of these spatiotemporal dynamics, the generated potential behavioral transfer network can depict the possible behavioral paths of giant pandas in different environments and reveal how their behavioral choices are affected by environmental factors and emotional changes.

[0043] Step S530: Perform multimodal decision modeling processing according to the potential behavior transfer network, generate a neural-behavioral coupling strategy set by jointly modeling visual attention focus, hormone level and exercise energy consumption, and learning the mapping relationship from environmental features to action strategies; In this process, the neural-behavioral coupling strategy set used can jointly analyze the giant panda's sensory input, internal physiological state and behavioral output. By modeling the focus of visual attention, it is possible to simulate how giant pandas choose to observe, focus on and respond to certain visual stimuli in a specific environment; changes in hormone levels (such as physiological changes such as appetite and anxiety) affect the giant panda's behavioral choices; and exercise energy consumption is directly related to the giant panda's activity and behavior duration. By learning the mapping relationship from environmental features to action strategies, the giant panda's behavioral strategy can be adaptively adjusted according to environmental changes, thereby generating behavioral predictions that are more in line with actual situations. The final generated neural-behavioral coupling strategy set provides multi-dimensional support for behavioral prediction and decision-making, and can take into account the giant panda's complex physiological, psychological and environmental interaction factors.

[0044] Step S540: Perform dynamic trajectory synthesis processing according to the neural-behavioral coupling strategy set, apply differential game theory to establish an individual-environment game model and solve the safe reachable set to obtain the behavior trajectory deduction result.

[0045] It is understandable that this process provides an optimization framework for the panda's behavioral path selection, especially when environmental changes or emotional fluctuations affect behavioral decisions, the game model can simulate how the panda can choose the optimal movement path while ensuring safety. The generated behavioral trajectory deduction results not only accurately predict the panda's action trajectory, but also take into account the dynamic changes in different environments and emotional states, improving the accuracy and practicality of the prediction.

[0046] Further, step S600 includes step S610 to step S640.

[0047] Step S610: Perform spatiotemporal semantic segmentation processing according to the behavior trajectory deduction results, and divide the continuous trajectory into behavior units that conform to the activity rhythm characteristics of the giant panda through a spectral clustering algorithm combined with terrain data to obtain a set of behavior semantic segments; It can be understood that spectral clustering is a technique for clustering analysis by constructing the spectrum of a graph, which can effectively process nonlinear and complex spatiotemporal data. In this step, the spectral clustering algorithm combines the giant panda's behavioral trajectory data and terrain data, and identifies continuous behavioral patterns in similar environments or states by calculating the similarity between trajectory data points. By performing spectral analysis on the spatiotemporal distribution of the giant panda's behavioral trajectory, the trajectory can be divided into several units with similar behavioral patterns, and these units reflect a series of behavioral activities of the giant panda in different time periods. Secondly, combined with the activity rhythm characteristics of the giant panda, that is, its behavioral pattern changes in different time periods (such as circadian activity rhythm, seasonal behavioral changes, etc.), it is possible to more accurately divide the representative behavioral units. For example, the foraging behavior of the giant panda is usually closely related to certain specific environmental conditions (such as bamboo growth areas or near water sources). The spectral clustering algorithm can segment the trajectory based on these environmental characteristics and identify the different behavioral stages of the giant panda in a specific environment. In this way, each segment in the trajectory can reflect the behavior pattern of the giant panda in a specific time and space dimension. This step can transform complex behavior trajectory data into behavior units with clear meanings through spatiotemporal semantic segmentation, providing more structured data support for subsequent behavior analysis and prediction.

[0048] Step S620: Perform pattern association mining based on the behavior semantic segment set, apply the tensor decomposition algorithm to fuse the three-dimensional features of time, space, and environment, extract the common patterns across behavior units through tensor decomposition with non-negative constraints, and obtain the behavior pattern feature tensor; It should be noted that, first, the tensor decomposition algorithm is used to perform pattern mining on the set of behavioral semantic fragments. Tensor decomposition is a high-dimensional data decomposition method that can extract relevant information from multidimensional data (such as time, space, and environment) and reveal the potential correlation between different dimensions. In this step, each behavior unit in the behavioral semantic fragment set is regarded as an element of a tensor, and each dimension of the tensor represents the time, space, and environment characteristics of the behavior. Through tensor decomposition with non-negative constraints, common patterns across behavioral units can be effectively extracted from these high-dimensional data, avoiding the influence of negative values ​​in the data and maintaining the positive characteristics in practical applications.

[0049] Through this processing, the tensor decomposition algorithm can not only reveal the correlation between different behavioral units, but also identify the behavioral patterns of giant pandas under time, space and environmental conditions. For example, the behavior patterns of giant pandas in different geographical regions may vary significantly, but their basic activity patterns (such as foraging, resting, exploring, etc.) will show certain commonalities in the time series. Tensor decomposition can extract these common patterns and generate behavioral pattern feature tensors, thereby providing a more refined pattern representation for subsequent behavioral analysis and prediction.

[0050] Step S630: Perform risk situation assessment processing according to the behavior pattern feature tensor, use fuzzy cognitive graph to model the correlation between human activity interference, natural enemy appearance probability and behavior abnormality, solve the risk propagation path through particle swarm optimization algorithm, and obtain the ecological risk heat map; Specifically, the fuzzy cognitive graph is a knowledge representation method based on graph structure, which describes the causal relationship between variables through nodes and edges. In this step, the fuzzy cognitive graph can link external environmental factors (such as human activities disturbing the habitat, the frequency of natural enemies, etc.) with the abnormal behavior of giant pandas (such as escape behavior, aggressive behavior or changes in activity range). By establishing this association model, the potential threat of different environmental factors to the behavior of giant pandas is quantified, and the possible response of giant pandas when encountering threats is predicted. Next, the particle swarm optimization algorithm is used to solve the risk propagation path. Particle swarm optimization is an optimization algorithm based on swarm intelligence that can simulate the process of particles searching for the optimal solution in the search space. In this step, the particle swarm optimization algorithm can efficiently solve the propagation path of ecological risks, and find the risk diffusion path and potential high-risk areas by simulating how different environmental risk factors (such as the appearance of natural enemies and the impact of human activities) affect the behavior of giant pandas.

[0051] Finally, the ecological risk heat map generated by the above method combines the risk transmission path and the behavioral pattern characteristics of giant pandas to form a graphical output showing the risk concentration area. This heat map can intuitively show the ecological risk distribution of giant pandas under different environmental conditions.

[0052] Step S640: Perform multimodal report generation processing based on the ecological risk heat map, adopt template adaptation technology based on knowledge graph, build structured narrative logic through semantic role labeling and event extraction, and generate a visual behavior monitoring report in combination with three-dimensional terrain rendering.

[0053] First, template adaptation technology based on knowledge graph is used to generate the framework and structure of the report. Knowledge graph is a graphical structure that represents entities and their relationships through nodes and edges, and is used to represent domain knowledge. In the report generation process, knowledge graph can help to organize various information in a structured manner, such as giant pandas' behavior patterns, environmental conditions, and ecological risk assessment results. Through template adaptation technology, the content and presentation form of the report can be dynamically adjusted according to different monitoring needs and data types to ensure the personalization and pertinence of the report content. For example, in different risk assessment scenarios, the report template can display the dynamic changes in giant pandas' behavior as needed, or highlight high-risk points in specific areas.

[0054] Next, semantic role labeling and event extraction techniques are applied to extract specific risk information from the ecological risk heat map and convert it into a structured narrative. This process uses automated technology to identify the core elements of the event, such as human activities, threats from natural enemies, abnormal behavior, etc., and annotates the relationship and role of these elements in the entire ecosystem. Through semantic role labeling, the semantic relationship between each entity and behavior can be clearly identified, such as "human activities have led to changes in the giant panda's habitat", and a corresponding description can be made; while event extraction is used to extract specific events from ecological risk data, such as "frequent natural enemy activities have led to changes in the behavior patterns of giant pandas", forming a complete structured narrative logic to ensure that the content of the report is not only accurate but also easy to understand.

[0055] Finally, 3D terrain rendering is used to visualize geographic data to more intuitively display the giant panda behavior monitoring results and the spatial distribution of ecological risks. Through 3D terrain rendering technology, the report can dynamically display the giant panda's behavior trajectory, high-risk areas, and the specific distribution of ecological threats in maps or charts. Combined with environmental data, the 3D view can not only show the giant panda's habitat, but also highlight the factors that affect its activities, such as the distribution of food resources and natural enemies.

[0056] Embodiment 2:

[0057] like Figure 2 As shown, this embodiment provides a wild giant panda behavior monitoring system based on artificial intelligence, the system includes: An acquisition module 901 is used to acquire an original data set of a target giant panda, where the original data set is video information, geographic location information, and environmental information of the same wild giant panda collected from multiple monitoring locations; Extraction module 902, used to perform feature extraction processing based on video information and environmental information, extract the action, posture and micro-expression information of the target giant panda, and analyze the feature change trend in different time periods and different scenes in combination with environmental conditions to obtain a feature set; The conversion module 903 is used to perform numerical processing according to the feature set, and obtain a virtual individual numerical representation by quantifying the behavior characteristics, health characteristics and emotional characteristics into an activity intensity index, a body balance index and an emotional fluctuation index respectively; Modeling module 904 is used to perform modeling processing according to the virtual individual numerical representation, and construct a giant panda movement function model by simulating the skeletal movement posture and combining the influence of health and emotional state on individual behavior selection; The deduction module 905 is used to deduce the behavior trajectory according to the giant panda's motor function model and geographical location information, and to simulate the dynamic behavior selection process of the giant panda in different scenarios using the time series model and the Markov decision process to obtain the behavior trajectory deduction result; Output module 906 generates a behavior monitoring report of the target giant panda based on the behavior trajectory deduction results.

[0058] In some embodiments disclosed in the present application, the extraction module 902 includes: The first extraction unit is used to perform motion feature extraction processing according to the video information, and sparsely decompose the motion features in the video by using a sparse coding algorithm, extract key frames and motion patterns of the giant panda's motion trajectory to construct a motion feature dictionary, and obtain a motion feature set; The second extraction unit is used to perform posture feature extraction processing according to the video information, and perform dimension reduction and feature decomposition on the posture information in the video by applying a non-negative matrix decomposition algorithm, extract the posture changes in different scenes, and obtain a posture feature set; The third extraction unit is used to perform micro-expression feature extraction processing based on the video information, extract local features of the ear shaking and eye expression changes in the giant panda's facial area through a local binary pattern algorithm, analyze the emotional state in combination with a preset typical facial structure and expression pattern, and obtain a micro-expression feature set; The fourth extraction unit is used to perform feature integration processing based on the action feature set, posture feature set and micro-expression feature set. By combining the action, posture and micro-expression features with environmental information, the impact of environmental factors on the behavior of giant pandas is analyzed, and a comprehensive feature set is obtained through feature weighting and normalization processing.

[0059] In some embodiments disclosed in this application, the conversion module 903 includes: The first conversion unit is used to perform behavior timing alignment processing according to the action features in the feature set, perform nonlinear alignment on action sequences of unequal lengths across monitoring points through a dynamic time warping algorithm, and introduce ambient light intensity as a time expansion penalty factor to calculate an activity intensity index; The second conversion unit is used to perform posture space embedding processing according to the posture features in the feature set, map the skeleton key points extracted from the multi-view video to the low-dimensional manifold space by using an isometric feature mapping algorithm, and obtain the posture balance index by calculating the covariance matrix of the principal curvature change rate and the ground contact pressure; The third conversion unit is used to perform emotion coupling analysis and processing according to the micro-expression features in the feature set, apply fuzzy cognitive graph to model the nonlinear relationship between micro-expression electromyographic signals, environmental temperature and humidity, and eating behavior, determine the phase change threshold of the emotional state by calculating the Lyapunov exponent spectrum of the attractor domain, and obtain the emotion fluctuation index; The fourth conversion unit is used to perform multimodal fusion processing according to the activity intensity index, the body balance index and the mood fluctuation index, perform Pareto frontier search on all indexes, and generate a virtual individual numerical representation by defining a fitness function including an energy consumption function and a social contact probability.

[0060] In some embodiments disclosed in the present application, the modeling module 904 includes: The first modeling unit is used to perform skeletal dynamics modeling processing according to the virtual individual numerical representation, and adopt differential algebraic equations to describe joint motion constraints to obtain a rigid-flexible coupling skeletal model; The second modeling unit is used to integrate the health status according to the rigid-flexible coupled skeleton model, establish a metabolic energy-exercise efficiency conversion model using the stochastic optimal control theory, describe the probability density evolution of health indicators through the Fock-Planck equation, and obtain a set of physiological constraint exercise strategies; The third modeling unit is used to perform emotion-behavior coupling processing according to the physiological constraint movement strategy set, and obtain an emotion enhancement decision tree by modeling the influence intensity of emotion fluctuations on path selection; The fourth modeling unit is used to perform multi-scale fusion processing according to the emotion enhancement decision tree to obtain a giant panda movement function model.

[0061] In some embodiments disclosed in this application, the deduction module 905 includes: The first deduction unit is used to perform terrain manifold modeling based on geographic location information, using a diffusion mapping algorithm to fuse slope, bamboo density, and stream distribution data, and generate a terrain accessibility map by calculating a terrain similarity matrix and combining joint torque constraints; The second deduction unit is used to discretize the spatiotemporal state according to the terrain accessibility map, establish a behavioral event flow model using the non-uniform Poisson process, and model the self-motivation effect of food discovery events on the movement path through the Hawkes process to obtain the potential behavior transfer network; The third deduction unit is used to perform multimodal decision modeling based on the potential behavior transfer network, and generates a neural-behavioral coupling strategy set by jointly modeling visual attention focus, hormone levels and motor energy consumption, and learning the mapping relationship from environmental features to action strategies; The fourth deduction unit is used to perform dynamic trajectory synthesis processing according to the neural-behavioral coupling strategy set, apply differential game theory to establish an individual-environment game model and solve the safe reachable set to obtain the behavior trajectory deduction result.

[0062] Embodiment 3:

[0063] Corresponding to the above method embodiment, this embodiment also provides a wild giant panda behavior monitoring device based on artificial intelligence. The wild giant panda behavior monitoring device based on artificial intelligence described below and the wild giant panda behavior monitoring method based on artificial intelligence described above can be referenced to each other.

[0064] Figure 3 FIG. 8 is a block diagram of a wild giant panda behavior monitoring device 800 based on artificial intelligence according to an exemplary embodiment. Figure 3 As shown, the wild giant panda behavior monitoring device 800 based on artificial intelligence may include: a processor 801, a memory 802. The wild giant panda behavior monitoring device 800 based on artificial intelligence may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0065] Wherein, the processor 801 is used to control the overall operation of the wild giant panda behavior monitoring device 800 based on artificial intelligence to complete all or part of the steps in the above-mentioned wild giant panda behavior monitoring method based on artificial intelligence. The memory 802 is used to store various types of data to support the operation of the wild giant panda behavior monitoring device 800 based on artificial intelligence, and these data may include, for example, instructions for any application or method for operating on the wild giant panda behavior monitoring device 800 based on artificial intelligence, and application-related data, such as contact data, messages sent and received, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the wild giant panda behavior monitoring device 800 based on artificial intelligence and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0066] In an exemplary embodiment, an artificial intelligence-based wild giant panda behavior monitoring device 800 can be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned artificial intelligence-based wild giant panda behavior monitoring method.

[0067] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned method for monitoring wild giant panda behavior based on artificial intelligence are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by a processor 801 of a wild giant panda behavior monitoring device 800 based on artificial intelligence to complete the above-mentioned method for monitoring wild giant panda behavior based on artificial intelligence.

[0068] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for monitoring the behavior of wild giant pandas based on artificial intelligence, characterized in that: include: Acquire an original data set of a target giant panda, wherein the original data set is video information, geographic location information, and environmental information of the same wild giant panda collected from multiple monitoring locations; Performing feature extraction processing according to the video information and the environmental information, extracting the action, posture and micro-expression information of the target giant panda, and analyzing the feature change trends in different time periods and different scenes in combination with the environmental conditions, to obtain a feature set; Numerical processing is performed according to the feature set, and the behavior characteristics, health characteristics and emotional characteristics are quantified into activity intensity index, body balance index and emotional fluctuation index respectively to obtain a virtual individual numerical representation; Modeling is performed according to the virtual individual numerical representation, and a giant panda movement function model is constructed by simulating the skeletal movement posture and combining the influence of health and emotional state on individual behavior selection; Deducing the behavior trajectory according to the giant panda's motor function model and the geographic location information, simulating the dynamic behavior selection process of the giant panda in different scenarios using a time series model and a Markov decision process, and obtaining the behavior trajectory deduction result; A behavior monitoring report of the target giant panda is generated based on the behavior trajectory deduction results.

2. The method for monitoring the behavior of giant pandas in the wild based on artificial intelligence according to claim 1, characterized in that: Perform feature extraction processing based on the video information and the environmental information, extract the action, posture and micro-expression information of the target giant panda, and analyze the feature change trends in different time periods and different scenes in combination with the environmental conditions to obtain a feature set, including: Performing motion feature extraction processing according to the video information, performing sparse decomposition on the motion features in the video by using a sparse coding algorithm, extracting key frames and motion patterns of the giant panda's motion trajectory to construct a motion feature dictionary, and obtaining a motion feature set; Performing posture feature extraction processing according to the video information, performing dimension reduction and feature decomposition on the posture information in the video by applying a non-negative matrix decomposition algorithm, extracting its posture changes in different scenes, and obtaining a posture feature set; Performing micro-expression feature extraction processing based on the video information, extracting local features of ear shaking and eye expression changes in the facial area of ​​the giant panda using a local binary pattern algorithm, analyzing the emotional state in combination with a preset typical facial structure and expression pattern, and obtaining a micro-expression feature set; Feature integration processing is performed based on the action feature set, the posture feature set and the micro-expression feature set. The influence of environmental factors on the behavior of giant pandas is analyzed by combining the action, posture and micro-expression features with environmental information. A comprehensive feature set is obtained through feature weighting and normalization processing.

3. The method for monitoring wild giant panda behavior based on artificial intelligence according to claim 1, characterized in that: According to the feature set, numerical processing is performed, and the behavior characteristics, health characteristics and emotional characteristics are quantified into activity intensity index, body balance index and emotional fluctuation index respectively, so as to obtain the virtual individual numerical representation, including: Performing behavior timing alignment processing according to the action features in the feature set, performing nonlinear alignment on action sequences of unequal lengths across monitoring points through a dynamic time warping algorithm, and introducing ambient light intensity as a time expansion penalty factor to calculate an activity intensity index; Performing posture space embedding processing according to the posture features in the feature set, using an isometric feature mapping algorithm to map the skeleton key points extracted from the multi-view video to a low-dimensional manifold space, and obtaining a posture balance index by calculating the covariance matrix of the principal curvature change rate and the ground contact pressure; Emotional coupling analysis is performed based on the micro-expression features in the feature set, the nonlinear relationship between micro-expression electromyographic signals, environmental temperature and humidity, and eating behavior is modeled using fuzzy cognitive graphs, the phase transition threshold of the emotional state is determined by calculating the Lyapunov exponent spectrum of the attractor domain, and the emotional fluctuation index is obtained; Multimodal fusion processing is performed according to the activity intensity index, the body balance index and the mood fluctuation index, a Pareto frontier search is performed on all indexes, and a virtual individual numerical representation is generated by defining a fitness function including an energy consumption function and a social contact probability.

4. The method for monitoring wild giant panda behavior based on artificial intelligence according to claim 1, characterized in that: Modeling is performed based on the virtual individual numerical representation, and by simulating the skeletal movement posture and combining the influence of health and emotional state on individual behavior selection, a giant panda movement function model is constructed, including: Performing skeletal dynamics modeling processing according to the virtual individual numerical representation, using differential algebraic equations to describe joint motion constraints, and obtaining a rigid-flexible coupling skeletal model; According to the rigid-flexible coupled skeleton model, health status integration processing is performed, and a metabolic energy-exercise efficiency conversion model is established by applying stochastic optimal control theory. The probability density evolution of health indicators is described by the Focke-Planck equation to obtain a set of physiological constraint exercise strategies; Performing emotion-behavior coupling processing according to the physiological constraint movement strategy set, and obtaining an emotion enhancement decision tree by modeling the influence intensity of emotion fluctuations on path selection; Multi-scale fusion processing is performed according to the emotion enhancement decision tree to obtain a giant panda motor function model.

5. The method for monitoring wild giant panda behavior based on artificial intelligence according to claim 1, characterized in that: The behavior trajectory is deduced according to the giant panda's motor function model and the geographic location information, and the dynamic behavior selection process of the giant panda in different scenarios is simulated using a time series model and a Markov decision process to obtain the behavior trajectory deduction results, including: Performing terrain manifold modeling according to the geographic location information, using a diffusion mapping algorithm to fuse slope, bamboo density and stream distribution data, and generating a terrain accessibility map by calculating a terrain similarity matrix and combining joint torque constraints; According to the terrain accessibility map, the spatiotemporal state is discretized, a behavior event flow model is established by applying a non-uniform Poisson process, and the self-motivation effect of food discovery events on the movement path is modeled by a Hawkes process to obtain a potential behavior transfer network; Performing multimodal decision modeling processing according to the potential behavior transfer network, generating a neural-behavioral coupling strategy set by jointly modeling visual attention focus, hormone level and exercise energy consumption, and learning the mapping relationship from environmental features to action strategies; Dynamic trajectory synthesis processing is performed according to the neural-behavioral coupling strategy set, and the individual-environment game model is established by applying differential game theory and solving the safe reachable set to obtain the behavior trajectory deduction result.

6. A wild giant panda behavior monitoring system based on artificial intelligence, characterized in that: include: An acquisition module is used to acquire an original data set of a target giant panda, wherein the original data set is video information, geographic location information, and environmental information of the same wild giant panda collected from multiple monitoring locations; An extraction module is used to perform feature extraction processing according to the video information and the environmental information, extract the action, posture and micro-expression information of the target giant panda, and analyze the feature change trends in different time periods and different scenes in combination with the environmental conditions to obtain a feature set; A conversion module, used for performing numerical processing according to the feature set, and obtaining a virtual individual numerical representation by quantifying the behavior characteristics, health characteristics and emotional characteristics into an activity intensity index, a body balance index and an emotional fluctuation index respectively; A modeling module is used to perform modeling processing according to the virtual individual numerical representation, and construct a giant panda movement function model by simulating the skeletal movement posture and combining the influence of health and emotional state on individual behavior selection; A deduction module is used to deduce the behavior trajectory according to the giant panda's motor function model and the geographic location information, and to simulate the dynamic behavior selection process of the giant panda in different scenarios by using a time series model and a Markov decision process to obtain a behavior trajectory deduction result; The output module generates a behavior monitoring report of the target giant panda based on the behavior trajectory deduction results.

7. The wild giant panda behavior monitoring system based on artificial intelligence according to claim 6 is characterized in that: The extraction module comprises: A first extraction unit is used to perform motion feature extraction processing according to the video information, and sparsely decompose the motion features in the video by using a sparse coding algorithm, extract key frames and motion patterns of the giant panda's motion trajectory to construct a motion feature dictionary, and obtain a motion feature set; A second extraction unit is used to perform posture feature extraction processing according to the video information, and perform dimension reduction and feature decomposition on the posture information in the video by applying a non-negative matrix decomposition algorithm, extract the posture changes in different scenes, and obtain a posture feature set; A third extraction unit is used to perform micro-expression feature extraction processing based on the video information, extract local features of ear shaking and eye expression changes in the giant panda's facial area through a local binary pattern algorithm, analyze the emotional state in combination with a preset typical facial structure and expression pattern, and obtain a micro-expression feature set; The fourth extraction unit is used to perform feature integration processing based on the action feature set, the posture feature set and the micro-expression feature set, analyze the impact of environmental factors on the behavior of giant pandas by combining the action, posture and micro-expression features with environmental information, and obtain a comprehensive feature set through feature weighting and normalization processing.

8. The wild giant panda behavior monitoring system based on artificial intelligence according to claim 6 is characterized in that: The conversion module comprises: A first conversion unit is used to perform behavior timing alignment processing according to the action features in the feature set, perform nonlinear alignment on action sequences of unequal lengths across monitoring points through a dynamic time warping algorithm, and introduce ambient light intensity as a time expansion penalty factor to calculate an activity intensity index; A second conversion unit is used to perform posture space embedding processing according to the posture features in the feature set, map the skeleton key points extracted from the multi-view video to the low-dimensional manifold space by using an isometric feature mapping algorithm, and obtain a posture balance index by calculating the covariance matrix of the principal curvature change rate and the ground contact pressure; A third conversion unit is used to perform emotion coupling analysis and processing according to the micro-expression features in the feature set, use fuzzy cognitive graphs to model the nonlinear relationship between micro-expression electromyographic signals, environmental temperature and humidity, and eating behavior, determine the phase change threshold of the emotional state by calculating the Lyapunov exponent spectrum of the attractor domain, and obtain the emotion fluctuation index; The fourth conversion unit is used to perform multimodal fusion processing according to the activity intensity index, the body balance index and the mood fluctuation index, perform Pareto frontier search on all indexes, and generate a virtual individual numerical representation by defining a fitness function including an energy consumption function and a social contact probability.

9. The wild giant panda behavior monitoring system based on artificial intelligence according to claim 6 is characterized in that: The modeling module includes: A first modeling unit is used to perform skeletal dynamics modeling processing according to the virtual individual numerical representation, using differential algebraic equations to describe joint motion constraints to obtain a rigid-flexible coupling skeletal model; The second modeling unit is used to perform health status integration processing according to the rigid-flexible coupled skeleton model, establish a metabolic energy-exercise efficiency conversion model by applying the stochastic optimal control theory, describe the probability density evolution of health indicators by the Focke-Planck equation, and obtain a set of physiological constraint exercise strategies; A third modeling unit is used to perform emotion-behavior coupling processing according to the physiological constraint movement strategy set, and obtain an emotion enhancement decision tree by modeling the influence intensity of emotion fluctuations on path selection; The fourth modeling unit is used to perform multi-scale fusion processing according to the emotion enhancement decision tree to obtain a giant panda motor function model.

10. The wild giant panda behavior monitoring system based on artificial intelligence according to claim 6 is characterized in that: The deduction module includes: A first deduction unit is used to perform terrain manifold modeling according to the geographical location information, fuse the slope, bamboo density and stream distribution data using a diffusion mapping algorithm, and generate a terrain accessibility map by calculating a terrain similarity matrix and combining joint torque constraints; The second deduction unit is used to discretize the spatiotemporal state according to the terrain accessibility map, establish a behavior event flow model by applying a non-uniform Poisson process, and model the self-motivation effect of food discovery events on the movement path by using a Hawkes process to obtain a potential behavior transfer network; A third deduction unit is used to perform multimodal decision modeling processing according to the potential behavior transfer network, and generate a neural-behavioral coupling strategy set by jointly modeling visual attention focus, hormone level and exercise energy consumption, and learning the mapping relationship from environmental features to action strategies; The fourth deduction unit is used to perform dynamic trajectory synthesis processing according to the neural-behavioral coupling strategy set, apply differential game theory to establish an individual-environment game model and solve the safe reachable set to obtain the behavior trajectory deduction result.

Images