Monitoring and early warning method and system for wildlife nature reserves

Abstract

The present invention provides a monitoring and early warning method and system for wildlife nature reserves. First, a multimodal ecological monitoring data stream containing animal activity trajectory data, habitat environmental parameter sequences, and bio-voiceprint signal sets is collected. Next, the animal activity trajectory data is dynamically matched with the baseline threat pattern to generate a first ecological association network feature; the migration path fragments are extracted and combined with the environmental parameter change gradient, and the second ecological association network feature is derived through a model. Subsequently, the two features are cross-modally fused to generate a multi-level threat coupling vector, a map is constructed based on the bio-voiceprint and superimposed with the coupling vector for analysis, and a real-time threat level spectrum is output through adaptive algorithm calibration. Finally, based on the mapping relationship between the real-time threat level spectrum and the preset strategy library, the corresponding ecological protection intervention instruction set is activated to achieve comprehensive analysis of multimodal data, accurate real-time monitoring and early warning, and intelligent response, effectively protecting the ecology of wildlife nature reserves.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a monitoring and early warning method and system for wildlife nature reserves. Background Art

[0002] In the field of monitoring and early warning of wildlife nature reserves, traditional technologies mainly focus on single-dimensional monitoring.

[0003] Early monitoring methods relied heavily on regular manual patrols, which not only consumed significant manpower, material resources, and time, but also limited the scope of monitoring, making it difficult to provide a comprehensive and real-time understanding of the ecological status of nature reserves. With technological advancements, some automated monitoring equipment has been introduced, but most of these devices only monitor a single factor.

[0004] For example, some monitoring systems focus solely on measuring habitat parameters, such as temperature, humidity, and light intensity sensors. While these systems can capture certain environmental data, they are unable to effectively correlate this information with behavioral changes in wildlife. Consequently, even if an anomaly in an environmental parameter is detected, it is difficult to determine whether it poses an actual threat to wildlife, as well as the extent and scope of the threat. Summary of the Invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a monitoring and early warning method for a wildlife nature reserve, the method comprising:

[0006] Collecting multimodal ecological monitoring data streams within the target wildlife nature reserve, wherein the multimodal ecological monitoring data streams include animal activity trajectory data, habitat environment parameter sequences, and bio-voiceprint signal sets;

[0007] Dynamically matching the animal activity trajectory data with a baseline threat pattern in a predefined ecological threat indicator library to generate a first ecological association network feature, wherein the first ecological association network feature characterizes the spatiotemporal coupling relationship between the individual behavior of the target animal and the potential threat scenario;

[0008] Extracting migration path segments from the animal activity trajectory data, combining them with the change gradient of the habitat environmental parameter sequence, and deriving a second ecological association network feature through a threat pattern evolution model, wherein the second ecological association network feature reflects the synergistic effect strength between abnormal fluctuations in migration paths and sudden changes in environmental parameters;

[0009] Performing cross-modal feature fusion on the first ecological association network feature and the second ecological association network feature to generate a multi-level threat coupling vector;

[0010] Based on the frequency domain characteristic distribution of the biometric voiceprint signal set, a voiceprint anomaly detection map is constructed, and the voiceprint anomaly detection map is spatially superimposed and analyzed with the multi-level threat coupling vector. After generating a comprehensive threat situation matrix, an adaptive weight allocation algorithm is used to dynamically calibrate the threshold of the comprehensive threat situation matrix and output a real-time threat level spectrum;

[0011] According to the mapping relationship between the real-time threat level spectrum and the preset emergency response strategy library, the corresponding level of ecological protection intervention instruction set is activated.

[0012] On the other hand, an embodiment of the present invention also provides a monitoring and early warning system for a wildlife nature reserve, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0013] Based on the above aspects, the embodiments of the present application greatly improve the comprehensiveness, accuracy and timeliness of nature reserve monitoring and early warning. Specifically, the collection of multimodal ecological monitoring data streams covers animal activity trajectory data, habitat environmental parameter sequences and bio-voice signal sets. This multi-dimensional data acquisition method comprehensively and meticulously reflects the operating status of the protected area ecosystem. Compared with the traditional monitoring method of a single data source, multimodal data provides a rich information foundation.

[0014] By dynamically matching animal activity trajectory data with a predefined ecological threat indicator library, the generated first ecological association network features can accurately capture the spatiotemporal coupling relationship between the individual behavior of target animals and potential threat scenarios. This not only breaks the limitation of traditional monitoring that only focuses on a single factor, but also provides real-time insights into the dynamic changes of ecological threats in time and space dimensions, making monitoring and early warning more forward-looking and targeted.

[0015] By extracting migration path fragments and combining them with the gradient of changes in habitat environmental parameters, the characteristics of the second ecological association network are derived, reflecting the intensity of the synergistic effect of abnormal fluctuations in migration paths and sudden changes in environmental parameters, revealing the complex relationships between different elements in the ecosystem, and helping to deeply understand the intrinsic driving mechanism of ecological threats. Compared with traditional methods, it can better explore potential ecological risk factors.

[0016] Cross-modal feature fusion generates multi-level threat coupling vectors, effectively integrating key information from different modal data and avoiding the bias and limitations inherent in single-modal data. Multi-level feature fusion provides a more comprehensive and representative representation for subsequent threat analysis, significantly enhancing the overall understanding of ecological threats.

[0017] Based on the biological voiceprint signal set, a voiceprint anomaly detection map is constructed, and spatial overlay analysis is performed with multi-level threat coupling vectors. The generated comprehensive threat situation matrix comprehensively presents the ecological threat status within the protected area. The adaptive weight allocation algorithm further performs dynamic threshold calibration on the matrix. The output real-time threat level spectrum is more accurate and flexible, and can adapt to the dynamic changes of the ecological environment in a timely manner, providing a reliable basis for protection decisions.

[0018] Finally, based on the mapping relationship between the real-time threat level spectrum and the pre-set emergency response strategy library, the corresponding level of ecological protection intervention instruction set is activated, achieving a seamless connection between monitoring and early warning and protection actions. This greatly improves the efficiency and effectiveness of nature reserves in responding to ecological threats. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of the execution flow of the wildlife nature reserve monitoring and early warning method provided by an embodiment of the present invention.

[0020] Figure 2 It is a schematic diagram of the hardware architecture of a wildlife nature reserve monitoring and early warning system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0021] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a method for monitoring and early warning of wildlife nature reserves provided by an embodiment of the present invention. The method for monitoring and early warning of wildlife nature reserves is introduced in detail below.

[0022] Step S110 , collecting a multimodal ecological monitoring data stream within a target wildlife nature reserve, wherein the multimodal ecological monitoring data stream includes animal activity trajectory data, a habitat environment parameter sequence, and a bio-voiceprint signal set.

[0023] In this embodiment, in a large wildlife nature reserve, such as a giant panda nature reserve located in a mountainous area, various types of monitoring equipment can be widely deployed in the reserve to collect multimodal ecological monitoring data streams. For the collection of animal activity trajectory data, collars or trackers with positioning functions can be installed in key areas of the reserve, and these devices are worn by giant pandas. These positioning devices can record the geographic location information of the giant panda at a certain time interval, such as every 15 minutes, and transmit the data to the data center via satellite communication or a local wireless transmission network. In this way, as time goes by, continuous animal activity trajectory data can be formed, showing information such as the giant panda's activity range and movement path in the reserve.

[0024] The collection of habitat environmental parameter sequences is achieved through a network of multiple sensors. Temperature and humidity sensors, light intensity sensors, and soil moisture sensors have been deployed in various ecological zones within the reserve, such as bamboo forests, near streams, and on hillsides. The temperature and humidity sensors monitor the ambient temperature and humidity in real time. For example, during the hottest summer months, they can accurately record temperature fluctuations within the bamboo forest, from a relatively cool 20 degrees Celsius in the morning to potentially soaring to 30 degrees Celsius in the afternoon. Light intensity sensors measure the intensity of sunlight at different times of day, which is important for understanding the preferred lighting conditions in the giant panda's habitat. Soil moisture sensors measure soil moisture content, which increases significantly during the rainy season and decreases during the dry season. The data collected by these sensors is arranged chronologically to form a habitat environmental parameter sequence.

[0025] The collection of bio-voiceprint signal sets relies on highly sensitive sound collection equipment. Within the protected area, this embodiment installs these sound collection devices on trees or hidden locations to ensure that the sounds made by living things can be captured to the greatest extent possible. For example, giant pandas may make various sounds to communicate or express emotions. After these sounds are captured by the collection equipment, they form a bio-voiceprint signal set. At the same time, the calls of other wild animals such as birds and small mammals are also collected. These sound signals contain different characteristics such as frequency, duration, and intensity, which together constitute a rich bio-voiceprint signal set.

[0026] Step S120 , dynamically matching the animal activity trajectory data with a baseline threat pattern in a predefined ecological threat indicator library to generate a first ecological association network feature, wherein the first ecological association network feature characterizes the spatiotemporal coupling relationship between the individual behavior of the target animal and the potential threat scenario.

[0027] Taking the giant panda nature reserve mentioned above as an example, there are multiple baseline threat patterns in the predefined ecological threat indicator library. Suppose that one of the baseline threat patterns is a threat related to human activities, such as humans illegally entering the core activity area of giant pandas. When the collected giant panda activity trajectory data shows that a giant panda suddenly changed its regular activity path, moving from the originally safe depths of the bamboo forest to the edge of the reserve, and in the process the activity trajectory became irregular and the speed also changed. At this time, this embodiment first processes the animal activity trajectory data, first performs noise filtering on it, eliminates abnormal data points caused by occasional equipment failures or environmental interference, and then performs trajectory interpolation processing to ensure the continuity of the trajectory and generate a standardized activity trajectory sequence. For the habitat environment parameter sequence, time alignment is performed to ensure consistency with the animal activity trajectory data on a time scale, and at the same time, outliers are corrected to generate a continuous environmental parameter surface.

[0028] The standardized activity trajectory sequence is mapped to the continuous environmental parameter surface in a spatiotemporal grid. For example, the protected area is divided into small grid cells. The size of each cell can be determined according to actual needs and data accuracy, for example, 1 square kilometer is a cell. Within each grid cell, the behavior density distribution and environmental parameter gradient are extracted. For example, in the grid cells where giant pandas are active frequently, the behavior density distribution is high. If the environmental parameter gradient of this area shows a sudden increase in noise level or abnormal changes in temperature and humidity caused by surrounding human activities,

[0029] The threat pattern matching engine then compares the behavioral density distribution and environmental parameter gradients with the spatiotemporal constraints of the baseline threat model. If the temporal and spatial characteristics of giant panda activity near areas of frequent human activity match the spatiotemporal constraints of the baseline threat model, where illegal human intrusion causes behavioral changes, a threat match confidence score is generated for each grid cell. For example, the threat match confidence score is high in several grid cells near the edge of the protected area, while it is low in grid cells farther from human activity areas. Based on the threat match confidence scores for all grid cells, a three-dimensional threat heat map is constructed as the first ecological association network feature. In this three-dimensional threat heat map, height represents the degree of threat match confidence, and different colors represent different threat levels. This intuitively illustrates the spatiotemporal coupling between individual giant panda behaviors and potential threat scenarios, such as which areas of giant pandas are most vulnerable to human activity and the severity of the threat.

[0030] Step S130: extract the migration path segments from the animal activity trajectory data, combine them with the change gradient of the habitat environmental parameter sequence, and derive the second ecological association network characteristics through the threat pattern evolution model. The second ecological association network characteristics reflect the synergistic effect strength between abnormal fluctuations in migration paths and sudden changes in environmental parameters.

[0031] Continuing with the example of giant panda nature reserves, this embodiment identifies path turning points and rest periods within giant panda trajectory data. For example, giant pandas may migrate from lower-elevation bamboo forests to higher-elevation areas each spring. During this process, there are path turning points when they enter from one bamboo forest to another, and they may rest near water sources. Using this information, this embodiment segments the migration path into subsegments with continuous motion vectors.

[0032] Curvature analysis and speed variation detection are performed on each migration path sub-segment. If the curvature of the giant panda's movement path suddenly increases within a certain migration path sub-segment, such as a relatively straight path becoming curved, or if there is a significant variation in speed, such as a sudden acceleration or deceleration from normal slow movement, this embodiment marks these abnormal path morphological features. At the same time, combined with the gradient of the habitat environmental parameter sequence, for example, in the bamboo forest near the giant panda's migration path, due to abnormal climate, the temperature suddenly drops, affecting the growth of bamboo, and the soil moisture also undergoes sudden changes due to changes in precipitation.

[0033] A causal correlation analysis is performed between the abnormal path morphological features and the environmental parameter mutation events within the corresponding time window. If it is found that the abnormal changes in the giant panda's path are correlated with the environmental parameter mutations in time and space, for example, in an area where the temperature drops sharply, the giant panda changes its migration path or speed, this embodiment calculates the path-environment coupling coefficient. Assuming that in a certain sub-segment of the migration path, the environmental parameter mutation is very drastic, and the giant panda's path also shows obvious abnormal changes, then this path-environment coupling coefficient will be higher. A dynamic threat propagation chain is constructed based on the path-environment coupling coefficients of all migration path sub-segments as the second ecological association network feature. This dynamic threat propagation chain can show the extent of the impact of environmental parameter mutations on giant panda migration on different migration paths, and how this impact is propagated between different path segments, reflecting the intensity of the synergistic effect between abnormal fluctuations in the migration path and environmental parameter mutations.

[0034] Step S140, cross-modal feature fusion is performed on the first ecological association network feature and the second ecological association network feature to generate a multi-level threat coupling vector, and based on the frequency domain feature distribution of the biological voiceprint signal set, a voiceprint anomaly detection map is constructed, and the voiceprint anomaly detection map is spatially superimposed and analyzed with the multi-level threat coupling vector. After generating a comprehensive threat situation matrix, the comprehensive threat situation matrix is dynamically threshold calibrated using an adaptive weight allocation algorithm to output a real-time threat level spectrum.

[0035] In the monitoring scenario of the Giant Panda Nature Reserve, principal component dimensionality reduction was first performed on the first ecological association network feature (the three-dimensional threat heat map). During this process, this embodiment extracted the core features that best represent threat information from the three-dimensional threat heat map, such as those representing representative grid cells with high threat match confidence, and converted them into core threat space vectors.

[0036] This embodiment converts the second ecological association network feature (dynamic threat propagation chain) into a spatiotemporal propagation probability matrix. This matrix represents the probability of a threat propagating along a migratory path at different temporal and spatial locations. A tensor product operation is then performed on the spatiotemporal propagation probability matrix and the core threat space vector to generate a multi-level threat coupling vector. This vector integrates information from two different modalities: the spatiotemporal coupling relationship between individual animal behavior and potential threat scenarios, and the strength of the synergistic effect between abnormal fluctuations in migratory paths and sudden changes in environmental parameters.

[0037] This embodiment performs frame segmentation and windowing on the biometric voiceprint signal set. For example, the continuous sound signal is segmented into frames of a specific length (e.g., 0.1 seconds per frame) and an appropriate window function is applied to improve the accuracy of the spectral analysis. The mel-frequency cepstral coefficients and harmonic energy ratios of each frame are then extracted. These features are then combined and identified using an abnormal voiceprint classifier. For example, if the mel-frequency cepstral coefficients and harmonic energy ratios of a particular sound significantly differ from those of a normal giant panda or other biological sound, it is classified as an abnormal voiceprint. A voiceprint anomaly probability distribution map is then output, with different regions in the map indicating the probability of an abnormal voiceprint. Anomalous voiceprint events are then spatially clustered based on a sound source localization algorithm. For example, if multiple abnormal voiceprint events are detected in a specific area of a protected area, these events are grouped together, and a location-marked voiceprint anomaly detection map is generated.

[0038] The location markers of the voiceprint anomaly detection map are aligned with the spatial coordinates of the multi-level threat coupling vector. For example, if the voiceprint anomaly detection map indicates a voiceprint anomaly in the northeastern region of the protected area, and the multi-level threat coupling vector also shows a high threat coupling value in this area, accurate spatial alignment is performed. The overlap percentage and spatial correlation index between the voiceprint anomaly and threat coupling regions are then calculated to generate a comprehensive threat situation matrix. This comprehensive threat situation matrix integrates threat information from multiple aspects, including animal activity trajectories, habitat environment, and biological voiceprints.

[0039] Next, an adaptive weight allocation algorithm is used to dynamically calibrate thresholds for the comprehensive threat situation matrix. Based on the threat level distribution patterns in the historical threat event database, baseline threshold curves are established for each threat dimension (such as animal behavior threat, environmental threat, and voiceprint threat). For example, for the animal behavior threat dimension, if similar instances of abnormal animal behavior in the past have generally resulted in a medium threat level, this baseline threshold curve is used as a basis for establishing this. A sliding time window algorithm is used to calculate the cumulative intensity of each threat dimension in the current comprehensive threat situation matrix in real time. For example, the cumulative intensity of animal behavior threat, environmental threat, and voiceprint threat dimensions is calculated over a short time window (e.g., one hour). The cumulative intensity values are mapped onto the baseline threshold curve using a nonlinear interpolation method, outputting a dynamically adjusted real-time threat level spectrum. This spectrum accurately reflects the current threat level within the protected area, such as low, medium, or high threat, based on the current situation.

[0040] Step S150: activating an ecological protection intervention instruction set of a corresponding level according to a mapping relationship between the real-time threat level spectrum and a preset emergency response strategy library.

[0041] For example, in the context of a giant panda nature reserve, suppose the real-time threat level spectrum indicates a high threat level. In the pre-configured emergency response strategy library, a corresponding set of ecological protection intervention instructions is provided for this high threat level. First, the weight distribution of each threat dimension in the real-time threat level spectrum is analyzed. For example, if the weight of the animal behavior threat dimension is high, it indicates that the animal may be in immediate danger. Meanwhile, if the weights of the environmental threat dimension and the voiceprint threat dimension are also high, it may indicate the presence of multiple threat factors acting together.

[0042] Based on this weight distribution, the multi-level response protocols in the preset emergency response strategy library are matched. For situations with high threat levels and high animal behavior threat weights, response protocols such as emergency animal rescue and enhanced perimeter patrols may be matched. Based on the spatial coverage and duration of the threat level, drone patrol path planning, ecological corridor blockade instructions, and manual patrol priority queues are generated. If the high threat level covers a large area and lasts for a long time, drone patrol paths will be planned over a larger area, blocking potentially dangerous ecological corridors, and increasing the priority of manual patrols so that inspectors can focus on high-threat areas.

[0043] Finally, the command set is synchronized to the protected area management terminal and mobile law enforcement equipment via a low-latency communication network. Upon receiving the command, the protected area management terminal can dispatch management resources across the entire protected area, such as deploying additional staff and equipment to high-threat areas. Upon receiving the command, the mobile law enforcement equipment can then perform specific actions based on the instructions, such as setting up blockade signs in ecological corridors and conducting patrols along planned routes. This allows for timely and effective response to high-threat situations within the protected area, safeguarding the living environment and safety of wildlife like giant pandas.

[0044] Based on the above steps, the embodiments of the present application greatly improve the comprehensiveness, accuracy, and timeliness of nature reserve monitoring and early warning. Specifically, the collection of multimodal ecological monitoring data streams covers animal activity trajectory data, habitat environmental parameter sequences, and bio-voiceprint signal sets. This multi-dimensional data acquisition method comprehensively and meticulously reflects the operating status of the protected area ecosystem. Compared with the traditional monitoring method of a single data source, multimodal data provides a rich information foundation.

[0045] By dynamically matching animal activity trajectory data with a predefined ecological threat indicator library, the generated first ecological association network features can accurately capture the spatiotemporal coupling relationship between the individual behavior of target animals and potential threat scenarios. This not only breaks the limitation of traditional monitoring that only focuses on a single factor, but also provides real-time insights into the dynamic changes of ecological threats in time and space dimensions, making monitoring and early warning more forward-looking and targeted.

[0046] By extracting migration path fragments and combining them with the gradient of changes in habitat environmental parameters, the characteristics of the second ecological association network are derived, reflecting the intensity of the synergistic effect of abnormal fluctuations in migration paths and sudden changes in environmental parameters, revealing the complex relationships between different elements in the ecosystem, and helping to deeply understand the intrinsic driving mechanism of ecological threats. Compared with traditional methods, it can better explore potential ecological risk factors.

[0047] Cross-modal feature fusion generates multi-level threat coupling vectors, effectively integrating key information from different modal data and avoiding the bias and limitations inherent in single-modal data. Multi-level feature fusion provides a more comprehensive and representative representation for subsequent threat analysis, significantly enhancing the overall understanding of ecological threats.

[0048] Based on the biological voiceprint signal set, a voiceprint anomaly detection map is constructed, and spatial overlay analysis is performed with multi-level threat coupling vectors. The generated comprehensive threat situation matrix comprehensively presents the ecological threat status within the protected area. The adaptive weight allocation algorithm further performs dynamic threshold calibration on the matrix. The output real-time threat level spectrum is more accurate and flexible, and can adapt to the dynamic changes of the ecological environment in a timely manner, providing a reliable basis for protection decisions.

[0049] Finally, based on the mapping relationship between the real-time threat level spectrum and the pre-set emergency response strategy library, the corresponding level of ecological protection intervention instruction set is activated, achieving a seamless connection between monitoring and early warning and protection actions. This greatly improves the efficiency and effectiveness of nature reserves in responding to ecological threats.

[0050] In a possible implementation, before step S120, the method further includes:

[0051] Noise filtering and trajectory interpolation processing are performed on the animal activity trajectory data to generate a standardized activity trajectory sequence.

[0052] The habitat environmental parameter sequence is time-aligned and outlier corrected to generate a continuous environmental parameter surface.

[0053] In this embodiment, for the animal activity trajectory data, the original data may contain noise due to the influence of the device itself or environmental factors. For example, the positioning device may be affected by electromagnetic interference from the complex terrain of the mountainous area, resulting in small jumps or deviations in the recorded location information. If this noise data is not processed, it will affect the accuracy of subsequent analysis. Therefore, noise filtering is required, and specific algorithms, such as filtering algorithms based on statistical models, are used to identify and remove data points that obviously deviate from the normal trajectory fluctuation range. At the same time, due to some unavoidable factors, such as temporary loss of connection of the device or interruption of signal transmission, the trajectory data may be discontinuous. At this time, trajectory interpolation processing is particularly important. This embodiment adopts a suitable interpolation method, such as spline interpolation, to supplement reasonable values at the position of missing data according to the previous and next trends of the trajectory data, thereby generating a standardized activity trajectory sequence.

[0054] The habitat environmental parameter sequence also faces similar problems. Different sensors may collect data at different frequencies, which requires time alignment operations. For example, the temperature and humidity sensor may collect data every 10 minutes, while the light intensity sensor may collect data every 15 minutes. In order to ensure that these data can accurately correspond on a time scale, this embodiment uses a unified time base, such as a time interval of 10 minutes, to adjust the data collected by the light intensity sensor and align it with the data of the temperature and humidity sensor in time. In addition, the sensor may have abnormal values due to failures or extreme environmental factors. For example, in heavy rain, the soil moisture sensor may briefly give an excessively high humidity value due to being soaked in water. This embodiment identifies and corrects these abnormal values by comparing with the surrounding sensor data or based on statistical analysis methods of historical data, thereby generating a continuous environmental parameter surface.

[0055] Step S120 includes:

[0056] Step S121 , performing spatiotemporal grid mapping on the standardized activity trajectory sequence and the continuous environmental parameter surface, and extracting the behavior density distribution and environmental parameter gradient within the grid unit.

[0057] In step S122 , the threat pattern matching engine compares the behavior density distribution and the environmental parameter gradient with the spatiotemporal constraints of the baseline threat pattern to generate a threat matching confidence level for each grid cell.

[0058] Step S123 : constructing a three-dimensional threat heat distribution map based on the threat matching confidence of all grid cells as the first ecological association network feature.

[0059] When generating the first ecological association network feature, the standardized activity trajectory sequence is first mapped to the continuous environmental parameter surface in a spatiotemporal grid. The giant panda nature reserve is divided into numerous grid cells, each assuming a square area of 1 square kilometer. Within each grid cell, the behavioral density distribution of animal activity trajectories is calculated. For example, if a grid cell is frequented by giant pandas and stays there for a long time, then the behavioral density of this cell is high. At the same time, the environmental parameter gradient is analyzed, such as the change in temperature or humidity from one side of the grid cell to the other. If, within a grid cell, the temperature on the side closest to human activity is significantly higher than on the other side away from human activity, this indicates the presence of a certain environmental parameter gradient.

[0060] Next, the threat pattern matching engine compares the behavioral density distribution and environmental parameter gradients with the spatiotemporal constraints of the baseline threat pattern. The spatiotemporal constraints of the baseline threat pattern are predefined. For example, frequent human presence within a certain range of a giant panda's activity area during a certain time period is considered a threat pattern. If, within a grid cell, the animal behavior density distribution indicates frequent giant pandas near areas of high human activity, and the environmental parameter gradients also indicate environmental changes caused by human activity, such as increased noise levels, the threat match confidence for that grid cell is high. Finally, a three-dimensional threat heat map is constructed based on the threat match confidence of all grid cells. In this map, the x- and y-axes represent the geographic coordinates of the protected area, and the z-axis represents the threat match confidence. Different colors represent different threat levels. This intuitively illustrates the spatiotemporal coupling between individual animal behaviors and potential threat scenarios across the entire protected area.

[0061] In a possible implementation, step S130 includes:

[0062] Step S131 : identifying path turning points and dwell periods in the standardized activity trajectory sequence, and segmenting migration path sub-segments with continuous motion vectors.

[0063] Step S132 : performing curvature analysis and speed variation detection on each migration path sub-segment, and marking abnormal path morphological features.

[0064] Step S133 , performing causal correlation analysis between the abnormal path morphological features and the environmental parameter mutation events within the corresponding time window, and calculating the path-environment coupling coefficient.

[0065] Step S134 : constructing a dynamic threat propagation chain based on the path-environment coupling coefficients of all migration path subsegments as the second ecological association network feature.

[0066] When analyzing the activity trajectory data of giant pandas to obtain the second ecological association network characteristics, this embodiment starts with the standardized activity trajectory sequence. Identifying the turning points and stay periods of the path is the key first step. For example, when a giant panda is migrating, when it crosses from one bamboo forest to another, its movement direction may change significantly. This point is the path turning point. Near the water source, the giant panda may stay for a long time to drink water and rest. This is the stay period. Based on this information, this embodiment can segment the migration path sub-segments with continuous motion vectors.

[0067] Performing curvature analysis and speed variation detection on each migration path sub-segment is an important means to gain an in-depth understanding of the giant panda's migration behavior. Assume that in a certain migration path sub-segment, under normal circumstances, the giant panda's movement path should be a relatively smooth curve, but if due to certain factors, such as encountering a steep hillside or encountering interference from human activities, its path curvature may suddenly increase and become curved. At the same time, speed variation detection can also detect some abnormal situations. For example, the giant panda's speed is relatively stable during normal migration, but if it suddenly speeds up or slows down in a certain section of the road, this may indicate that there is an abnormality in the surrounding environment. This embodiment marks these abnormal path morphological features for subsequent analysis.

[0068] Causally analyzing the relationship between unusual path morphological features and environmental parameter mutations within the corresponding time window is a key step in constructing the second ecological association network feature. For example, if this embodiment discovers that a giant panda's speed suddenly slows and its path curvature increases within a certain migration path subsegment, and that the environmental parameters in that area also undergo a sudden change during this time period, such as a sudden drop in temperature that alters bamboo growth and reduces the giant panda's food resources, this embodiment calculates the path-environment coupling coefficient by analyzing this temporal and spatial correlation. If the environmental parameter mutation in a specific migration path subsegment significantly affects the giant panda's migration path and speed, then this path-environment coupling coefficient will be high. Finally, a dynamic threat propagation chain is constructed based on the path-environment coupling coefficients for all migration path subsegments. This dynamic threat propagation chain can demonstrate how environmental parameter mutations affect the giant panda's migratory behavior along different migration paths and how these effects propagate between different path segments, thereby reflecting the strength of the synergistic effect between unusual migration path fluctuations and environmental parameter mutations.

[0069] And, step S140 includes:

[0070] Step S141 : performing principal component dimensionality reduction processing on the three-dimensional threat heat distribution map to extract a core threat space vector.

[0071] Step S142 : Convert the dynamic threat propagation chain into a spatiotemporal propagation probability matrix, and perform a tensor product operation with the core threat space vector to generate the multi-level threat coupling vector.

[0072] For the three-dimensional threat heat map in the first ecological association network feature, due to its high data dimensionality and potential for redundant information, this embodiment performs principal component dimensionality reduction (PCD). Using a PCA algorithm, the principal components that best represent threat information are identified, thereby extracting the core threat space vector. For example, in a three-dimensional threat heat map, multiple factors may influence the confidence level of threat matching, such as animal behavior density in different areas or variations in environmental parameters. However, PCD allows this embodiment to identify the key factor combination that best represents the overall threat profile and convert it into a core threat space vector.

[0073] This embodiment converts the dynamic threat propagation chain in the second ecological association network feature into a spatiotemporal propagation probability matrix. This matrix is constructed based on an analysis of the path-environment coupling coefficients of each migration path subsegment in the dynamic threat propagation chain. For example, this embodiment calculates the probability of the threat propagating between these locations based on the path-environment coupling coefficients of each migration path subsegment at different temporal and spatial locations. If the path-environment coupling coefficient is high in a particular migration path subsegment, then the probability of the threat propagating from this subsegment to the adjacent subsegment is greater, and the corresponding position in the spatiotemporal propagation probability matrix will have a higher probability value.

[0074] Finally, a tensor product operation is performed on the spatiotemporal propagation probability matrix and the core threat space vector. This operation integrates information from two different modalities: the first, the spatiotemporal coupling relationship between individual animal behavior and potential threat scenarios, as reflected in the ecological association network; and the second, the synergistic effect between abnormal fluctuations in migration routes and sudden changes in environmental parameters, as reflected in the ecological association network. This tensor product operation generates a multi-level threat coupling vector. This multi-level threat coupling vector provides a more comprehensive and accurate reflection of the overall threat situation within the giant panda nature reserve, providing a stronger basis for subsequent threat assessment and response measures.

[0075] In a possible implementation, step S140 further includes:

[0076] Step S143 , performing frame-by-frame windowing processing on the biometric voiceprint signal set, and extracting the Mel-frequency cepstral coefficients and harmonic energy ratios of each frame signal.

[0077] Step S144: performing a joint discrimination on the Mel-frequency cepstral coefficients and the harmonic energy ratio through an abnormal voiceprint classifier, and outputting a voiceprint abnormality probability distribution diagram.

[0078] Step S145: spatially cluster abnormal voiceprint events based on a sound source localization algorithm to generate a voiceprint anomaly detection map with orientation marks.

[0079] Step S146: aligning the orientation mark of the voiceprint anomaly detection map with the spatial coordinates of the multi-level threat coupling vector.

[0080] Step S147: Calculate the overlapping area ratio and spatial correlation index of the voiceprint abnormal area and the threat coupling area to generate the comprehensive threat situation matrix.

[0081] In this embodiment, the bio-voiceprint signal is a continuous time series signal, such as various sound signals emitted by giant pandas, including sounds when foraging, sounds of communicating with companions, and the calls of other wild animals. Frame-segmenting and windowing processing is to divide this continuous signal into signal segments according to a certain time length, forming frame-by-frame signal segments. Assume that the length of each frame is set to 0.1 seconds. This time length is determined based on the vocal characteristics of wild animals such as giant pandas and the needs of signal analysis. Windowing processing is to multiply each frame signal by a specific window function, such as the Hanning window function, in order to reduce spectral leakage and improve the accuracy of spectral analysis.

[0082] After framing and windowing, the Mel-frequency cepstral coefficients and harmonic energy ratios are extracted for each frame. Mel-frequency cepstral coefficients are a spectral feature based on the human auditory perception that effectively describes the frequency characteristics of sound signals. For giant panda calls, different vocalizations correspond to different Mel-frequency cepstral coefficients. For example, the Mel-frequency cepstral coefficients for a giant panda's friendly calls and its calls when frightened differ significantly. The harmonic energy ratio reflects the energy ratio between the harmonic and fundamental components in a sound signal. Different organisms exhibit unique characteristics in their harmonic energy ratios when making sounds. Using a specific algorithm, the framed signal is calculated to accurately extract the Mel-frequency cepstral coefficients and harmonic energy ratio for each frame.

[0083] Next, the abnormal voiceprint classifier performs a joint discrimination of the Mel-frequency cepstral coefficients and harmonic energy ratios. This model is trained using a large amount of known normal and abnormal biological voiceprint sample data. It determines whether the current frame's sound signal is abnormal based on the input Mel-frequency cepstral coefficients and harmonic energy ratio features. For example, when a giant panda is in a normal state, the Mel-frequency cepstral coefficients and harmonic energy ratios of its calls fluctuate within a certain range. If these feature values corresponding to a particular frame's sound signal exceed the normal range, the abnormal voiceprint classifier will classify it as abnormal. After evaluating all frames in the biological voiceprint signal set, a voiceprint abnormality probability distribution map is output. This map is based on the geographical regions of the protected area, with each region corresponding to a voiceprint abnormality probability value. A higher probability value indicates a greater likelihood of an abnormal voiceprint in that region.

[0084] Then, anomalous soundprint events are spatially clustered based on a sound source localization algorithm. This algorithm determines the location of abnormal soundprint events by analyzing information such as the time and intensity differences between different sensors receiving abnormal soundprint signals. For example, multiple sound collection devices are distributed within a protected area. When an abnormal soundprint event occurs in a certain area, the time and intensity of the signals received by different collection devices will vary. These differences are used to calculate the approximate location of the abnormal soundprint event. These abnormal soundprint events with similar locations and characteristics are then spatially clustered and classified into the same category. Finally, a voiceprint anomaly detection map with azimuth markings is generated. This map not only shows which areas within the protected area have abnormal soundprint events, but also marks the azimuth information of these events. For example, there is a high probability of abnormal soundprint events in an area northeast of the protected area.

[0085] After constructing the voiceprint anomaly detection map, it is necessary to perform spatial overlay analysis with the multi-level threat coupling vector to generate a comprehensive threat situation matrix. First, the orientation mark of the voiceprint anomaly detection map is aligned with the spatial coordinates of the multi-level threat coupling vector. The multi-level threat coupling vector contains threat information obtained based on the analysis of animal activity trajectory data and habitat environmental parameter sequences, and its spatial coordinates correspond to different areas within the protected area. The orientation mark of the voiceprint anomaly detection map also points to a specific area within the protected area. Through precise geographic coordinate matching and other methods, the accurate spatial correspondence between the two is ensured. For example, the abnormal voiceprint area marked in the voiceprint anomaly detection map in the middle of the protected area must be accurately matched with the threat information corresponding to the central area in the multi-level threat coupling vector.

[0086] After registration, the overlap percentage and spatial correlation index of the voiceprint anomaly region and the threat coupling region are calculated. Voiceprint anomaly regions are areas with a high probability of being abnormal in the voiceprint anomaly detection map, while threat coupling regions are areas with high threat coupling values in the multi-level threat coupling vector. The overlap percentage is the ratio of the overlapping area of these two regions to their total area, reflecting the degree of spatial overlap between the voiceprint anomaly and other threat factors. The spatial correlation index is calculated using a more complex mathematical model that takes into account factors such as the spatial similarity and distance between the two regions. For example, if the voiceprint anomaly region and the threat coupling region largely overlap, the overlap percentage will be high. Similarly, if their spatial distribution shapes and positional relationships are very similar, the spatial correlation index will also be high. By calculating these two indicators, a comprehensive threat situation matrix is generated. This matrix combines the anomalies reflected by the bio-voiceprint signal with the threat situation analyzed based on the animal's movement trajectories and habitat environment, providing a more comprehensive picture of the overall threat situation within the protected area.

[0087] In a possible implementation, step S140 may further include:

[0088] Based on the threat level distribution pattern in the historical threat event database, a baseline threshold curve for each threat dimension is established.

[0089] The sliding time window algorithm is used to calculate the cumulative intensity value of each threat dimension in the current comprehensive threat situation matrix in real time.

[0090] The cumulative intensity value is mapped onto the baseline threshold curve by a nonlinear interpolation method, and a dynamically adjusted real-time threat level spectrum is output.

[0091] And, step S150 may include:

[0092] Step S151 , analyzing the weight distribution of each threat dimension in the real-time threat level spectrum, and matching the multi-level response protocol in the preset emergency response strategy library.

[0093] In step S152 , based on the spatial coverage and duration of the threat level, drone patrol path planning, ecological corridor blocking instructions, and manual inspection priority queues are generated.

[0094] Step S153: Synchronize the instruction set to the protected area management terminal and mobile law enforcement equipment via a low-latency communication network.

[0095] In this embodiment, the historical threat event database stores data related to various threat events that occurred in the giant panda nature reserve in the past. These data record in detail the threat situations and corresponding threat levels under different threat dimensions. The threat dimensions cover the previously analyzed animal behavior threat dimensions, habitat environment threat dimensions, and bio-voice threat dimensions. For example, in the animal behavior threat dimension, the database records events in which giant pandas change their normal activity patterns due to interference from human activities. When giant pandas frequently approach the edge of the reserve or move away from their normal habitat range, they are judged to have different levels of threat based on past experience, which may correspond to different degrees and frequencies of behavioral abnormalities from low to high. For the habitat environment threat dimension, factors such as abnormal temperature fluctuations, sudden changes in humidity, or vegetation destruction in past events and their corresponding threat levels are also recorded. The bio-voice threat dimension records the correlation between abnormal bio-voice events and actual threats, such as the threat levels corresponding to certain abnormal giant panda calls or the frightened calls of other wild animals in different situations.

[0096] Through in-depth analysis of this historical data, a baseline threshold curve is established for each threat dimension. For example, in this example, we might find that when a giant panda approaches the edge of a protected area more than a certain number of times in a month, the threat level gradually increases. Based on this pattern, we plot a curve with the number of approaches to the protected area on the horizontal axis and the threat level on the vertical axis. This serves as the baseline threshold curve for that threat dimension. Similar curves are constructed for other threat dimensions.

[0097] Next, a sliding time window algorithm is used to calculate the cumulative intensity value of each threat dimension in the current comprehensive threat situation matrix in real time. The sliding time window algorithm sets a specific time window length, such as a 24-hour time window. Within this time window, the data in the comprehensive threat situation matrix is continuously updated, and the cumulative intensity value of each threat dimension is calculated. For the animal behavior threat dimension, if the giant panda has multiple abnormal activity paths or an abnormal reduction in activity range within this 24-hour period, the threat intensity corresponding to these behaviors will be cumulatively calculated. In terms of the habitat environment threat dimension, if the temperature continues to rise or the humidity continues to fall beyond the normal range within this time window, the threat intensity brought about by these environmental changes will also be cumulatively calculated. The biological voiceprint threat dimension is no exception. If the frequency of abnormal voiceprint events is detected to increase or the abnormality of the abnormal voiceprint increases within 24 hours, its threat intensity will also be cumulatively calculated.

[0098] The cumulative intensity values are then mapped onto a baseline threshold curve using a nonlinear interpolation method, outputting a dynamically adjusted real-time threat level spectrum. This nonlinear interpolation method takes into account that the relationship between threat intensity and threat level is not a simple linear one. For example, when the cumulative intensity values of the animal behavior threat dimension are within a certain range, the threat level may slowly increase. However, when the cumulative intensity values exceed a certain threshold, the threat level may rise sharply. The calculated cumulative intensity values are then mapped onto the previously established baseline threshold curve using this nonlinear relationship to obtain the real-time threat level for each threat dimension. The real-time threat levels of each threat dimension are combined to form a real-time threat level spectrum. This spectrum reflects the overall threat level of the giant panda nature reserve in real time, reflecting the combined effects of various threat factors, such as whether the overall threat level is low, medium, or high.

[0099] When activating the corresponding level of ecological protection intervention instruction set based on the mapping relationship between the real-time threat level spectrum and the pre-set emergency response strategy library, the first step is to analyze the weight distribution of each threat dimension in the real-time threat level spectrum. The weights of different threat dimensions in the real-time threat level spectrum reflect the relative importance of each dimension within the current overall threat situation. For example, if the weight of the animal behavior threat dimension is high at a certain moment, this may mean that abnormal behavior of giant pandas is the main threat factor at present, perhaps because the giant pandas have experienced severe behavioral changes due to significant external disturbances. A low weight for the habitat environment threat dimension may indicate that the current environmental factors are relatively stable. A medium weight for the bio-voice threat dimension may indicate that although there are some abnormal bio-voice events, their contribution to the overall threat is not the most significant.

[0100] Based on this weight distribution, the multi-level response protocols in the preset emergency response strategy library are matched. The preset emergency response strategy library contains a variety of response strategies for different threat situations. If the weight of the animal behavior threat dimension is high and the overall threat level is high, the response protocols that may be matched include immediately dispatching professionals to observe the giant panda closely to check for injuries or other abnormalities, while increasing the intensity of investigations on surrounding human activities to prevent further interference. If the weight of the habitat environment threat dimension suddenly increases, for example, due to severe deterioration of the habitat environment due to extreme weather, the corresponding response protocol may be to initiate an emergency habitat restoration plan, such as emergency water replenishment or provision of temporary food resources.

[0101] Then, based on the spatial coverage and duration of the threat level, drone patrol path planning, ecological corridor blockade instructions, and manual patrol priority queues are generated. If the spatial coverage of the threat level is large, for example, covering most areas of the protected area, then the drone patrol path planning needs to cover a wider area to ensure comprehensive monitoring of the entire threatened area. For ecological corridors, if the threat level remains high and involves the ecological corridor area, an ecological corridor blockade instruction needs to be issued to prevent external interference factors from entering the core protected area through the corridor. At the same time, the manual patrol priority queue is determined based on the threat level and spatial coverage. Areas with high threat levels and large coverage are listed as priority areas for manual patrols, and more patrol personnel and resources are allocated for key patrols.

[0102] Finally, the command set is synchronized to the protected area management terminal and mobile law enforcement equipment via a low-latency communication network. This ensures that the command is transmitted quickly and accurately. After receiving the command set, the protected area management terminal can centrally dispatch resources within the protected area and coordinate the implementation of appropriate conservation intervention measures. After receiving the command set, the mobile law enforcement equipment can quickly take action according to the instructions, such as executing ecological corridor blockades or conducting manual inspections according to priority queues. This allows for timely and effective response to various threats within the protected area, safeguarding the living environment and ecological security of wildlife such as giant pandas.

[0103] In one possible implementation, the method further includes:

[0104] A digital twin model of the ecosystem of the target wildlife nature reserve is established to receive and map the multimodal ecological monitoring data stream in real time.

[0105] A threat evolution simulator is embedded in the digital twin model to predict the threat diffusion path within the future time window based on the current comprehensive threat situation matrix.

[0106] The execution sequence and resource allocation plan of the ecological protection intervention instruction set are optimized based on the prediction results.

[0107] The updating step of the digital twin model includes:

[0108] Collect ecological status feedback data after the actual implementation of intervention measures and calculate the threat suppression efficiency coefficient.

[0109] The threat suppression efficiency coefficient is input into the threat evolution simulator for back propagation training to update the prediction parameters of the threat diffusion path.

[0110] When the threat suppression efficiency coefficient is lower than a preset threshold, the secondary response mechanism is triggered and the cross-regional collaborative intervention instruction set is regenerated.

[0111] In this embodiment, the digital twin model is a comprehensive digital mapping of the giant panda nature reserve ecosystem, encompassing numerous elements within the reserve, including topography, vegetation distribution, water resources, and the distribution and activity patterns of various wildlife. The multimodal ecological monitoring data stream includes a wealth of information, including the aforementioned animal trajectory data, habitat environmental parameter sequences, and bio-soundprint signal sets. For example, for animal trajectory data, the corresponding module in the digital twin model accurately depicts the giant panda's movement paths within the reserve based on the received data, including their seasonal migration routes within the bamboo forest and their dwelling time at water sources. Data such as temperature, humidity, light intensity, and soil moisture in the habitat environmental parameter sequences are mapped to corresponding geographic regions in the digital twin model, thereby reflecting the environmental conditions of each region. The bio-soundprint signal set is also reflected in the digital twin model. For example, when receiving a specific giant panda call signal, the model can determine the approximate location of the call and its possible meaning (such as a warning or courtship) based on pre-analysis.

[0112] Next, a threat evolution simulator is embedded in the digital twin model. It predicts threat spread paths within a future time window based on the current comprehensive threat situation matrix. The comprehensive threat situation matrix integrates multiple threat information, including animal movement patterns, habitat conditions, and bio-voice patterns. The threat evolution simulator leverages this information and integrates the interrelationships between ecosystem components to perform simulations and predictions. For example, if the current comprehensive threat situation matrix indicates a high level of human activity threat (such as illegal intrusion or construction) near a particular giant panda habitat, and the habitat is also experiencing some degradation (such as bamboo loss due to pollution), the threat evolution simulator will use these factors and ecosystem interactions to predict the potential spread of this threat to surrounding areas over a period of time (e.g., within the next week). This simulator takes into account the movement patterns of giant pandas. For example, habitat threats may cause giant pandas to migrate to other areas, which could put pressure on the new area's ecology, further affecting the threat's spread path. Furthermore, the status of other wildlife, as reflected in bio-voice patterns, is taken into account. If other animals shift their range or behavior patterns in response to threats, this could also impact the spread of the threat.

[0113] Then, based on the prediction results, the execution sequence and resource allocation plan for the ecological protection intervention instruction set are optimized. If the prediction results indicate that the threat to a certain area will increase dramatically in the next few days, the execution sequence of the ecological protection intervention instruction set for that area needs to be advanced. For example, the inspection and maintenance of a potentially threatened ecological corridor originally planned for a week later may need to be adjusted to be carried out immediately or within a shorter timeframe based on the prediction results. In terms of resource allocation, if the threat in a certain area spreads rapidly and has a large impact, more resources will need to be allocated to that area. For example, the frequency of drone patrols in the area can be increased, more human inspectors can be dispatched, or more equipment can be invested in environmental remediation.

[0114] The first step in updating the digital twin model is to collect feedback data on the ecological status after the implementation of actual intervention measures. After implementing a series of ecological protection intervention instructions to address threats within the protected area, such as habitat restoration, strengthening human activity controls, or providing medical treatment for injured giant pandas, feedback data on various aspects of the ecosystem after these interventions is collected. This data includes the recovery of vegetation in the restored habitat, whether the giant pandas' behavior has returned to normal, and whether illegal human activities have been effectively curbed.

[0115] Next, the Threat Suppression Efficiency Coefficient is calculated. This coefficient is an important indicator for measuring the effectiveness of ecological conservation interventions in suppressing threats. For example, if a series of measures are implemented in a certain area to reduce the threat posed by human activities to giant pandas, a quantitative Threat Suppression Efficiency Coefficient is calculated by comparing multiple factors before and after the intervention, such as the frequency of illegal human activities, the recovery of giant pandas' behavior, and the degree of improvement in their habitat. If, after the intervention, illegal human activities have almost disappeared, giant pandas' activities have returned to normal, and the habitat has significantly improved, the Threat Suppression Efficiency Coefficient will be high. Conversely, if, after the intervention, significant illegal human activities persist, there is no significant improvement in giant pandas' behavior, and the environment has not been effectively improved, the Threat Suppression Efficiency Coefficient will be low.

[0116] The threat suppression efficiency coefficient is input into the Threat Evolution Simulator for backpropagation training, updating the predicted parameters for threat diffusion paths. When the Threat Evolution Simulator is initially constructed, its predicted parameters may contain certain errors or be inaccurate. By inputting the calculated threat suppression efficiency coefficient into backpropagation training, the simulator's parameters can be adjusted to make its predictions more accurate. For example, if the threat suppression efficiency coefficient in a particular area is low, it indicates that the simulator's previous understanding of the relationships between ecosystem components and threat transmission mechanisms in that area is skewed. Through backpropagation training, parameters such as the correlation weights between ecological components related to that area and the threat transmission rate can be adjusted to improve the accuracy of threat diffusion path predictions.

[0117] When the threat suppression efficiency coefficient falls below a preset threshold, the secondary response mechanism is triggered and a new set of cross-regional coordinated intervention instructions is generated. The preset threshold is a critical value set based on the protected area's ecological protection objectives and past experience. If the calculated threat suppression efficiency coefficient falls below this preset threshold, it means that the current intervention measures are ineffective and more forceful measures are needed. The secondary response mechanism involves cross-regional coordinated actions. For example, if the threat suppression effect is poor in a certain local area of the Giant Panda Nature Reserve, more resources may need to be allocated from the entire reserve or even from surrounding related areas. The regenerated cross-regional coordinated intervention instruction set may include coordinating with surrounding reserves to jointly strengthen the protection of giant panda migration routes, jointly carrying out large-scale habitat restoration projects, and uniformly dispatching more law enforcement forces to crack down on illegal human activities, etc., to ensure that the ecosystem of the Giant Panda Nature Reserve can be effectively protected.

[0118] In one possible implementation, the method further includes:

[0119] Build a human activity monitoring network around the target wildlife nature reserve to collect traffic flow data, nighttime light intensity distribution and illegal intrusion alarm signals.

[0120] Human activity monitoring data are cross-validated with the comprehensive threat situation matrix to identify complex ecological threats induced by human factors.

[0121] Automatically generate law enforcement evidence collection clue packages and public warning information push strategies based on complex threat characteristics.

[0122] The step of cross-validating the human activity monitoring data with the comprehensive threat situation matrix to identify complex ecological threats induced by human factors includes:

[0123] Analyze the spatiotemporal coupling between the spatiotemporal trajectories of human activities and abnormal animal behavior events, and calculate the contribution factors of human interference.

[0124] Combined with the persistent changing trends of habitat environmental parameters, the long-term cumulative effects of human activities on ecological threats are assessed.

[0125] When the human interference contribution factor exceeds a critical value, the directional tracking and image acquisition functions of the high-precision video surveillance equipment are activated.

[0126] In this embodiment, traffic flow monitoring equipment, such as induction loops or video monitoring equipment, is installed on roads surrounding the Giant Panda Nature Reserve. These devices can accurately record the number and type of vehicles passing through at different times. Traffic flow data is crucial for understanding the potential impact of human activities on the reserve, as vehicle traffic can cause noise, exhaust emissions, and other impacts, and may also indicate the flow of people. Nighttime light intensity distribution is monitored using light intensity sensors located at various locations around the reserve. For example, in villages or human activity areas near the edge of the reserve, abnormally high light intensity at night may indicate illegal nighttime activities, such as poachers using strong lights for searching or illumination, or certain construction activities at night that generate strong light interference. Illegal intrusion alarm signals are obtained through fence sensors and infrared monitoring equipment installed at the boundary of the reserve. When a person or large animal crosses the fence or triggers the infrared monitoring, an illegal intrusion alarm signal is generated, which directly reflects the possibility of illegal entry into the reserve.

[0127] Human activity monitoring data is cross-validated with a comprehensive threat situation matrix to identify complex ecological threats induced by human factors. The comprehensive threat situation matrix integrates threat information from multiple sources, including animal movement trajectories, habitat environments, and bio-soundprints. First, the spatiotemporal coupling between human activity trajectories and unusual animal behavior events is analyzed to calculate the contribution factor of human interference. For example, if traffic flow monitoring equipment indicates an increase in vehicles entering a protected area along a certain road section during a certain period, and at the same time, giant pandas' movement trajectories within the protected area show movement away from that road section and unusual behavior, such as increased activity frequency and disruption of their regular foraging and resting patterns, the temporal and spatial correlation between these human activity trajectories and unusual animal behavior events is analyzed through precise temporal and spatial coordinate matching, and the contribution factor of human interference is calculated. If the correlation is very strong, such as if giant pandas exhibit noticeable unusual behavior shortly after an increase in vehicles, the contribution factor of human interference will be higher.

[0128] Then, combined with the persistent changing trends of habitat environmental parameters, the long-term cumulative effects of human activities on ecological threats are evaluated. For example, exhaust emissions caused by long-term traffic flow may lead to a decline in air quality around the protected area. By monitoring the persistent changing trends of habitat environmental parameters such as the concentration of pollutants in the air, the long-term impact of such human activities on the ecosystem can be evaluated. If, during years of monitoring, it is found that with the annual increase in traffic flow, the growth of vegetation around the protected area is inhibited and the soil quality declines, this indicates that human activities have a long-term cumulative effect on ecological threats. If, within a certain period of time, not only does traffic flow increase, but the nighttime light intensity also increases abnormally, and the giant panda's habitat environment continues to deteriorate, and abnormal animal behavior increases, this indicates that there may be a complex ecological threat induced by human factors.

[0129] When the human disturbance contribution factor exceeds a critical value, the directional tracking and image acquisition functions of high-precision video surveillance equipment are activated. This critical value is set based on long-term research on the protected area's ecosystem and past experience. For example, previous research indicates that a human disturbance contribution factor of 0.8 could pose a serious threat to the survival of giant pandas and the ecosystem. When the calculated human disturbance contribution factor exceeds this critical value, high-precision video surveillance equipment located at key locations within the protected area is activated. These devices have high resolution and precise directional tracking capabilities, allowing them to track and film specific areas or targets. For example, if human activity is concentrated near the core habitat of giant pandas and the human disturbance contribution factor exceeds the critical value, the video surveillance equipment will conduct directional tracking of the area, recording image information such as the movement paths and physical features of those entering the area, providing strong evidence for subsequent law enforcement and analysis.

[0130] Automatically generate law enforcement evidence packages and public warning information push strategies based on complex threat characteristics. Once a complex ecological threat induced by human factors is identified, a law enforcement evidence package is automatically generated based on the specific characteristics of the threat. For example, if an area with abnormally elevated nighttime light intensity overlaps with a giant panda's activity area, and combined with intrusion alarm signals, confirms unauthorized entry, the law enforcement evidence package will include light intensity data for that area, the time and location of the intrusion alarm, and relevant video surveillance images. This information can be directly provided to law enforcement personnel, enabling them to conduct rapid and accurate law enforcement investigations. Simultaneously, a public warning information push strategy is generated based on the complex threat characteristics. If the complex threat primarily stems from human activities in surrounding villages, such as excessive tourism development or agricultural activities threatening a protected area, the public warning information push strategy might include disseminating information to residents of surrounding villages about the importance of protecting giant pandas and the potential harm their activities may cause. This information could be disseminated via mobile phone text messages, community bulletin boards, or dedicated environmental awareness platforms to raise public awareness and reduce human-induced threats to the protected area's ecosystem.

[0131] In one possible implementation, the method further includes:

[0132] Edge computing devices are deployed at key ecological nodes in the target wildlife nature reserve to perform real-time feature extraction on the multimodal ecological monitoring data streams collected locally.

[0133] The feature extraction results of each node are encrypted and aggregated through the federated learning framework to update the parameters of the global threat detection model.

[0134] The updated model parameters are dynamically distributed to each edge computing device to achieve the collaborative evolution of distributed threat perception capabilities.

[0135] The updating steps of the federated learning framework include:

[0136] Design a differential privacy protection strategy to add geographic mask noise to sensitive animal location information.

[0137] The model parameter differential aggregation technology is used to eliminate the impact of skewed data distribution of edge nodes on the global model.

[0138] When a node device is detected to be abnormally offline, the model parameter rollback mechanism is activated and the computing tasks are reallocated to adjacent nodes.

[0139] In this embodiment, key ecological nodes include the main habitats of giant pandas, such as core bamboo forest areas, water sources, and important migration routes. Edge computing devices are deployed at these locations, for example, small, low-power edge computing devices are installed on multiple bamboo trees in the core bamboo forest area. These devices can perform real-time processing of locally collected multimodal ecological monitoring data streams. For animal activity trajectory data, edge computing devices can directly perform preliminary analysis of the raw location information transmitted by positioning devices to extract features such as the boundaries of the giant panda's range within the area and areas of frequent activity. Taking the activities of giant pandas in a bamboo forest as an example, edge computing devices can determine the bamboo forest sub-areas where giant pandas frequently move, and whether their range tends to expand or contract toward the edges of the bamboo forest based on location data over a period of time. For habitat environmental parameter sequences, edge computing devices can perform real-time analysis of data collected by local temperature and humidity sensors, soil moisture sensors, and other sensors to extract characteristics of changing environmental parameters. For example, edge computing devices near a water source can quickly determine short-term fluctuations in soil moisture around the water source, as well as long-term moisture trends associated with seasonal changes. For biometric voiceprint signal sets, edge computing devices can perform preliminary processing on the locally collected sound signals, extracting characteristics such as the frequency distribution of biometric voiceprints within a specific time period and the changes in vocalization frequency. For example, during the giant panda breeding season, edge computing devices can analyze whether changes in the frequency of giant panda calls in the area are related to breeding behavior.

[0140] The federated learning framework encrypts and aggregates the feature extraction results from each node to update the parameters of the global threat detection model. After multiple key ecological nodes within the protected area perform feature extraction on their own edge computing devices, the federated learning framework integrates these decentralized feature extraction results. The federated learning framework uses encryption technology to ensure data security and privacy. For example, each edge computing device encrypts the extracted feature results and sends them to a central node (the data remains encrypted throughout this process to prevent sensitive information from being leaked). At the central node, these encrypted feature results are aggregated. For example, suppose one node extracts a feature indicating a decrease in the range of giant pandas in a certain area, while another node extracts a feature indicating an abnormally high ambient temperature in the same area. Through encrypted aggregation within the federated learning framework, these different features are combined to update the parameters of the global threat detection model. The global threat detection model comprehensively assesses ecological threats within the protected area. Updated parameters enable more accurate threat identification and assessment. For example, as the features from each node are continuously aggregated and updated, the model can more accurately determine whether a decrease in the range of giant pandas and an abnormally high ambient temperature indicate a higher ecological threat, as well as the type and severity of the threat.

[0141] Updated model parameters are dynamically distributed to each edge computing device, enabling the co-evolution of distributed threat perception capabilities. After the parameters of the global threat detection model are updated, these new parameters are dynamically distributed to each edge computing device. For example, when the global threat detection model adjusts its recognition parameters for a particular threat pattern (e.g., the shrinking of giant panda habitat due to human activity) based on new aggregated features, these new parameters are sent to each edge computing device. Each edge computing device utilizes these new model parameters to more accurately identify threats when analyzing locally collected data. For example, an edge computing device in a bamboo forest area, after receiving the new model parameters, can more accurately determine whether threats are caused by human activity (e.g., deforestation of surrounding bamboo forests) when analyzing giant panda movement trajectory data, environmental parameter data, and biometric voiceprint data. This enables the co-evolution of distributed threat perception capabilities. The threat perception capabilities of each edge computing device continuously improve as the global model parameters are updated, and this improvement is fed back into the next round of feature extraction and global model updates, forming a virtuous cycle.

[0142] Regarding the update step of the federated learning framework, a differential privacy protection strategy is first designed to add geo-masking noise to sensitive animal location information. Because the location information of wild animals, such as giant pandas, is extremely sensitive, a differential privacy protection strategy is designed to protect its privacy while ensuring efficient data processing. For example, when transmitting the precise location information of giant pandas in their core habitat to other nodes or the central node in the federated learning framework, geo-masking noise is added. This geo-masking noise is not simply random, but is designed based on the geographical characteristics of the protected area and the animal's activity patterns. For example, based on the topography of the giant panda's habitat, the actual location is blurred within a certain range. If a giant panda's precise coordinates are located in a valley, geo-masking noise is added. During data transmission, other nodes receive a blurred location within that valley. This ensures the privacy of the panda's precise location while not affecting overall data analysis and model training.

[0143] Model parameter differential aggregation technology is used to eliminate the impact of skewed data distribution at edge nodes on the global model. Due to differences in geographical environment, ecological conditions, and other factors, the data collected by edge computing devices at different key ecological nodes within the protected area may be skewed. For example, edge computing devices near water sources may collect more data on changes in water-related environmental parameters, while edge computing devices in bamboo forests may collect more data on environmental parameters related to bamboo growth. This skewed data distribution can affect the accuracy of the global threat detection model. Model parameter differential aggregation technology compensates for this data distribution difference when aggregating feature extraction results from each node. Specifically, when calculating global model parameters, rather than simply averaging the parameters of each node, they are weighted based on the distribution characteristics of the data from each node. For example, nodes with less collected data but more critical data characteristics are given higher weights. This eliminates the impact of skewed data distribution at edge nodes on the global model, ensuring that the global threat detection model accurately reflects the ecological status of the entire protected area.

[0144] When a node device is detected as abnormally offline, a model parameter rollback mechanism is initiated and computing tasks are reassigned to neighboring nodes. During the actual operation of the protected area, node devices may experience abnormal offline conditions due to device failures, power supply issues, or other unexpected circumstances. For example, an edge computing device in a bamboo forest may suddenly go offline due to a power outage caused by severe weather. When this situation is detected, a model parameter rollback mechanism is initiated. This mechanism rolls back the parameters of the global threat detection model to a previously stable version to prevent model parameter updates from being incorrectly updated due to the node's offline state. Simultaneously, computing tasks are reassigned to neighboring nodes. For example, computing tasks in a bamboo forest area originally handled by an offline device can be reassigned to other edge computing devices in the surrounding bamboo forest area. Upon receiving the new computing tasks, these neighboring nodes will continue feature extraction and data processing using their own collected data and previously stored relevant data, ensuring continuous and stable ecological monitoring and threat detection throughout the protected area.

[0145] In one possible implementation, the method further includes:

[0146] Integrate the historical ecological restoration records and species reproduction data of the target wildlife nature reserves to construct an ecosystem resilience assessment matrix.

[0147] A multidimensional correlation analysis was performed between the real-time threat level spectrum and the ecosystem resilience assessment matrix to predict the ecological restoration potential under different intervention strategies.

[0148] The implementation intensity and effect period of the ecological protection intervention instruction set are dynamically adjusted according to the prediction results.

[0149] The step of performing a multidimensional correlation analysis between the real-time threat level spectrum and the ecosystem resilience assessment matrix to predict the ecological restoration potential under different intervention strategies includes:

[0150] A threat intensity-restoration efficiency response surface model was established to quantify the restoration rate of key ecosystem indicators under different intervention measures.

[0151] Simulate the ecological resilience transition path under the superposition of multiple rounds of intervention measures to identify the optimal combination of intervention strategies.

[0152] When the restoration rate is predicted to be lower than the degradation rate, an expert consultation request is triggered and a protected area zoning management plan is initiated.

[0153] In this example, historical ecological restoration records include detailed information on various past ecological restoration projects within the Giant Panda Nature Reserve. For example, restoration work on damaged bamboo forests includes the timing, scope, methods used (e.g., species and quantity of bamboo replanted, soil improvement measures), and evaluations of restoration effectiveness. Species reproduction data includes information on the reproduction of giant pandas and other related species, such as their breeding cycles, reproductive success rates, and cub survival rates, as well as data on population changes and reproduction of other wild animals. By integrating these data, an ecosystem resilience assessment matrix is constructed. This matrix assesses the resilience of the protected area's ecosystem from multiple dimensions, such as vegetation recovery capacity, species reproduction stability, and the ecosystem's resistance to external disturbances. For example, in the case of bamboo forests in the giant panda habitat, the matrix quantifies the bamboo's self-renewal capacity, the stability of its food supply for the giant pandas, and its resilience to natural disasters (e.g., fire, pests, and diseases).

[0154] A multidimensional correlation analysis was conducted between the real-time threat level spectrum and the ecosystem resilience assessment matrix to predict the ecological recovery potential under different intervention strategies. The real-time threat level spectrum reflects the various threats currently present within the protected area, including threats to animal behavior, habitat environment, and bio-voice patterns, taking into account the threat level. Multidimensional correlation analysis explores the relationship between threat level and ecosystem resilience from multiple dimensions. For example, if the real-time threat level spectrum indicates a high level of habitat threat, such as water pollution or bamboo forest degradation, while the ecosystem resilience assessment matrix indicates that the bamboo forest's self-renewal capacity is weak, it is necessary to analyze the effectiveness of different intervention strategies in improving this situation.

[0155] A threat intensity-restoration efficiency response surface model was developed to quantify the restoration rate of key ecosystem indicators for different interventions. Different interventions have varying effects on various ecological issues within the Giant Panda Nature Reserve. For example, if the intervention for bamboo forest degradation involves large-scale replanting of locally adapted bamboo species while simultaneously improving soil fertility, a threat intensity-restoration efficiency response surface model can be used to quantify the restoration rate of key ecosystem indicators, such as bamboo forest area restoration and bamboo quality improvement. This model comprehensively considers the relationship between threat intensity (such as the degree of bamboo forest degradation, including the proportion of bamboo population loss and poor growth) and restoration efficiency (such as the increase in bamboo forest area over a given period and the health of the new bamboo). Assuming that the bamboo forest degradation level reaches 30%, and the aforementioned intervention is implemented, after one year, the model calculates that the bamboo forest area increases by 10% and the proportion of healthy bamboo increases by 15%. This quantifies the restoration rate of the intervention. Similarly, for other ecological issues, such as water pollution control and restoration of other wildlife habitats within protected areas, this model can be used to quantify the restoration rate of key indicators of the corresponding ecosystems due to different intervention measures.

[0156] Simulating the transition paths of ecological resilience under the combined effects of multiple successive rounds of intervention measures can identify the optimal combination of intervention strategies. In practical conservation efforts, multiple rounds of intervention are often necessary. For example, in the comprehensive restoration of giant panda habitats, the first round might involve controlling water pollution, the second round would involve improving the growing environment of bamboo forests, and the third round would involve strengthening protections for giant panda reproduction. By simulating the transition paths of ecological resilience under the combined effects of these successive rounds of intervention, we can visualize how the resilience of the entire ecosystem gradually changes. For example, after the first round of water pollution control, while the direct impact on giant pandas may not be significant, there will be some improvement in the health of other water-dependent organisms and the entire ecosystem, resulting in a slight increase in ecosystem resilience. After the second round of improving the growing environment of bamboo forests, the giant pandas' food resources increase, habitat quality improves, and ecosystem resilience further increases. After the third round of strengthening breeding protections, the giant panda population's reproduction is more secure, and the stability and resilience of the entire ecosystem are further enhanced. Through this simulation, we can identify which combination of intervention strategies will achieve the greatest improvement in ecosystem resilience in the shortest time and with the least resource investment, thereby determining the optimal combination of intervention strategies.

[0157] When the predicted restoration rate falls below the degradation rate, an expert consultation request is triggered, and a protected area zoning management plan is initiated. If the above analysis and simulations reveal that, for a particular ecological issue, such as the continued degradation of bamboo forests in a certain area, the predicted restoration rate falls below the degradation rate despite intervention measures, this indicates a critical situation. In this case, an expert consultation request is triggered, convening experts in related fields such as giant panda conservation, ecological restoration, and forestry. These experts will draw on their expertise and experience to conduct an in-depth analysis of the current situation. Simultaneously, a protected area zoning management plan is initiated. For example, giant panda nature reserves could be divided into different zones based on ecological status and threat levels. Stricter management measures may be implemented in severely degraded areas, such as restricting human access, increasing ecosystem monitoring, and increasing resources for targeted restoration. This ensures the stability of the entire protected area ecosystem and effectively protects the habitat of giant pandas and other wildlife.

[0158] Dynamically adjust the implementation intensity and duration of the ecological protection intervention directive set based on the prediction results. If the prediction results indicate that a certain intervention measure has a positive impact on ecological restoration, for example, an intervention measure for bamboo forest restoration results in a faster restoration rate than the degradation rate and achieves the expected ecological restoration goal in a shorter period of time, then the implementation intensity of the intervention measure in the ecological protection intervention directive set can be appropriately reduced, or its duration can be shortened to shift resources to other areas where they are needed. Conversely, if the prediction results indicate that a certain intervention measure is ineffective, such as an intervention measure for water source pollution control that has been implemented for a period of time but has not significantly improved the water source pollution situation, then it is necessary to increase the implementation intensity of the intervention measure, extend its duration, or adjust the specific content of the intervention measure, such as changing the pollution control technology or method, to improve the protection and restoration of the ecosystem.

[0159] For example, in one possible implementation, the method further includes:

[0160] An intelligent sound and light deterrent device is deployed at the boundary of the target wildlife nature reserve, and the triggering logic of the intelligent sound and light deterrent device is dynamically bound to the real-time threat level spectrum.

[0161] When illegal intrusion or high-risk animal behavior is detected, directional acoustic interference and laser warning modes are activated according to the direction and level of the threat.

[0162] The spatiotemporal distribution characteristics of deterrence device triggering events are recorded to optimize the sensitivity parameters of subsequent early warning algorithms.

[0163] The control steps of the intelligent sound and light deterrent device include:

[0164] Establish a species-adaptive deterrence strategy library based on the sound-sensitive frequency bands and light response characteristics of different animal species.

[0165] Combined with real-time meteorological conditions and terrain shielding coefficient, the sound wave propagation angle and light intensity attenuation compensation parameters are dynamically adjusted.

[0166] When the deterrence trigger frequency in the same area exceeds a set frequency threshold, the intelligent sound and light deterrence device is controlled to switch to a continuous active monitoring mode and increase the image sampling rate.

[0167] In this embodiment, the boundaries of the target wildlife nature reserve serve as the first line of defense against external interference and illegal intrusion. Intelligent sound and light deterrent devices are deployed along these boundaries. These intelligent sound and light deterrent devices can obtain real-time threat information within the reserve, derived from a real-time threat level spectrum. This spectrum integrates multiple threat factors, including animal activity patterns, habitat parameters, and bio-soundprints, to accurately reflect the overall threat level within the reserve. For example, if animal activity patterns within the giant panda nature reserve indicate a large number of unidentified animals approaching the boundary, or if habitat parameters indicate an unusual source of interference near the boundary, or if a bio-soundprint signal set detects unusual animal calls in the border area, the real-time threat level spectrum will comprehensively assess these threats and assign a corresponding level. The intelligent sound and light deterrent devices adjust their trigger sensitivity based on this dynamically changing threat level. If the threat level is low, the device may be in a relatively low-sensitivity standby mode; as the threat level increases, the device is ready to trigger countermeasures.

[0168] When illegal intrusion or high-risk animal behavior is detected, directional acoustic interference and laser warning modes are activated based on the threat's direction and level. Various monitoring devices, such as infrared sensors and video surveillance equipment, are deployed at the protected area's perimeter to detect illegal intrusion. For example, when infrared sensors detect someone crossing the protected area's boundary, or when video surveillance equipment identifies unauthorized persons or vehicles entering the protected area, this is considered an illegal intrusion. Detection of high-risk animal behavior is also based on multimodal ecological monitoring data streams. For example, when animal activity trajectory data indicates that aggressive wildlife frequently approaches the boundary and exhibits unusually agitated behavior, this is considered high-risk animal behavior. Upon detection of these situations, the intelligent acoustic and visual deterrent device responds based on the threat's direction and level. If the threat originates from the northeast of the protected area, the device will direct the directional acoustic interference and laser warning modes toward the northeast. The assessment of the threat level also determines the intensity of the response. Lower-level threats may involve brief, low-intensity acoustic interference and gentle laser warnings. However, higher-level threats, such as large-scale illegal intrusions or the approach of extremely dangerous animals, trigger high-intensity directional acoustic interference and a powerful laser warning mode. Directional acoustic interference uses sound waves of a specific frequency and intensity to repel trespassers or deter high-risk animals. These frequencies and intensities are carefully designed to achieve deterrence without causing unnecessary harm to the protected area's inhabitants and environment. Laser warning mode uses a bright laser beam to warn trespassers or high-risk animals away from the protected area's boundaries.

[0169] Record the spatiotemporal distribution characteristics of deterrent device triggering events to optimize the sensitivity parameters of subsequent early warning algorithms. Each time an intelligent acoustic and visual deterrent device is triggered, its spatiotemporal distribution characteristics are recorded in detail. For example, a record may be made of a deterrent device triggering event occurring in the southwest region of the protected area boundary at a specific time (e.g., 10:00 AM on May 10, 2023). This is because an illegal intrusion was detected. Over time, a large number of such triggering event records accumulate. These records are important for optimizing the sensitivity parameters of subsequent early warning algorithms. Frequent deterrent device triggering events in a particular area may indicate a complex threat situation or that the current early warning algorithm's sensitivity is insufficient in that area. By analyzing these spatiotemporal distribution characteristics, the sensitivity parameters of the early warning algorithm can be adjusted. For example, if illegal intrusions frequently occur in a certain area at night, but the early warning algorithm fails to trigger the deterrent device in a timely manner, the records can be analyzed to determine whether the sensitivity of the nighttime early warning algorithm in that area needs to be increased, enabling the intelligent acoustic and visual deterrent device to respond to threats more promptly and accurately.

[0170] The control steps for the intelligent sound and light deterrent device begin with establishing a library of species-adaptive deterrent strategies based on the sound-sensitive frequency bands and light response characteristics of different animal species. The Giant Panda Nature Reserve is home to a variety of animal species, each with distinct sound and light sensitivity bands and response characteristics. For example, giant pandas may be more sensitive to certain low-frequency sound waves, while some small mammals may be more sensitive to high-frequency sound waves. Regarding light response characteristics, some nocturnal animals may have a strong avoidance reaction to sudden bright light, while diurnal animals may react differently to specific colors of light. Through research and experiments on these diverse animal species, a library of species-adaptive deterrent strategies was established. Within this library, appropriate sound and light deterrent strategies are developed for each animal. When a specific animal is detected approaching the reserve boundary and the deterrent device needs to be activated, the strategy from this library is used. For example, when a giant panda is found approaching the border and behaving abnormally, the system will select sound waves of appropriate frequency and intensity and lasers of specific color and intensity for warning based on the deterrence plan for giant pandas in the strategy library. This can achieve the deterrent purpose while avoiding causing excessive fright or harm to the giant panda.

[0171] The sound wave propagation angle and light intensity attenuation compensation parameters are dynamically adjusted based on real-time weather conditions and terrain obscuration. Weather conditions and terrain factors significantly impact the effectiveness of the acoustic and visual deterrent system. For example, windy weather conditions can affect the direction and distance of sound wave propagation. If the wind blows from within the protected area toward the boundary, the sound waves may be dispersed and unable to effectively reach the target area. In this case, the sound wave propagation angle must be dynamically adjusted based on wind direction and speed to ensure that the sound waves accurately reach the threat source. Similarly, terrain obscuration can affect the effectiveness of acoustic and visual deterrents. If there are obstacles such as mountains or tall trees at the boundary of the protected area, they can block the propagation of sound waves and light. The terrain obscuration coefficient is calculated through terrain measurement and analysis, and the light intensity attenuation compensation parameters are adjusted accordingly. In areas with significant obstruction, the laser light intensity needs to be increased to ensure that the warning effect can penetrate the obstruction and reach the target.

[0172] When the frequency of deterrent triggers in the same area exceeds a set threshold, the intelligent sound and light deterrent device switches to continuous active monitoring mode and increases the image sampling rate. Certain areas of a protected area may experience frequent threats, leading to excessively high deterrent triggering rates. For example, near the boundary of a protected area with frequent human activity, people may frequently intrude or small animals may frequently trigger the deterrent device. When this frequency exceeds the set threshold, the intelligent sound and light deterrent device switches to continuous active monitoring mode to better understand the situation in the area and enable effective management. In this mode, the device no longer passively waits for triggers but instead continuously monitors the area. Simultaneously, the image sampling rate is increased, allowing video surveillance equipment to capture images of the area more frequently. This facilitates a more detailed observation of activity in the area, determining whether there is a persistent threat source or whether further adjustments to the deterrent strategy are needed. For example, if it is determined that the device is being triggered by the frequent activity of a small animal nesting in the area, targeted measures can be taken based on this more detailed image information, such as adjusting the deterrent device's trigger sensitivity or modifying the deterrent method without compromising the animal's survival.

[0173] Figure 2 FIG. 1 shows the hardware structure of a wildlife nature reserve monitoring and early warning system 100 for implementing the above-mentioned wildlife nature reserve monitoring and early warning method provided by an embodiment of the present invention, as shown in FIG. Figure 2 As shown, the wildlife nature reserve monitoring and early warning system 100 may include a processor 110 , a machine-readable storage medium 120 , a bus 130 , and a communication unit 140 .

[0174] The machine-readable storage medium 120 can store data and / or instructions. In some embodiments, the machine-readable storage medium 120 can store data obtained from an external terminal. In some embodiments, the machine-readable storage medium 120 can store data and / or instructions used by the wildlife nature reserve monitoring and early warning system 100 to execute or use the exemplary methods described herein.

[0175] During the specific implementation process, one or more processors 110 execute computer executable instructions stored in the machine-readable storage medium 120, so that the processor 110 can execute the wildlife nature reserve monitoring and early warning method as described in the above method embodiment. The processor 110, the machine-readable storage medium 120 and the communication unit 140 are connected through the bus 130, and the processor 110 can be used to control the sending and receiving actions of the communication unit 140.

[0176] The specific implementation process of the processor 110 can be found in the various method embodiments executed by the above-mentioned wildlife nature reserve monitoring and early warning system 100. The implementation principles and technical effects are similar and will not be repeated here in this embodiment.

[0177] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above-mentioned monitoring and early warning method for wildlife nature reserves is implemented.

[0178] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A monitoring and early warning method for wildlife nature reserves, characterized in that: The method comprises: Collecting multimodal ecological monitoring data streams within the target wildlife nature reserve, wherein the multimodal ecological monitoring data streams include animal activity trajectory data, habitat environment parameter sequences, and bio-voiceprint signal sets; Dynamically matching the animal activity trajectory data with a baseline threat pattern in a predefined ecological threat indicator library to generate a first ecological association network feature, wherein the first ecological association network feature characterizes the spatiotemporal coupling relationship between the individual behavior of the target animal and the potential threat scenario; Extracting migration path segments from the animal activity trajectory data, combining them with the change gradient of the habitat environmental parameter sequence, and deriving a second ecological association network feature through a threat pattern evolution model, wherein the second ecological association network feature reflects the synergistic effect strength between abnormal fluctuations in migration paths and sudden changes in environmental parameters; Performing cross-modal feature fusion on the first ecological association network feature and the second ecological association network feature to generate a multi-level threat coupling vector; Based on the frequency domain characteristic distribution of the biometric voiceprint signal set, a voiceprint anomaly detection map is constructed, and the voiceprint anomaly detection map is spatially superimposed and analyzed with the multi-level threat coupling vector. After generating a comprehensive threat situation matrix, an adaptive weight allocation algorithm is used to dynamically calibrate the threshold of the comprehensive threat situation matrix and output a real-time threat level spectrum; According to the mapping relationship between the real-time threat level spectrum and the preset emergency response strategy library, the corresponding level of ecological protection intervention instruction set is activated.

2. The method for monitoring and early warning of wildlife nature reserves according to claim 1, characterized in that: Before the step of dynamically matching the animal activity trajectory data with a baseline threat pattern in a predefined ecological threat indicator library, the method further includes: performing noise filtering and trajectory interpolation processing on the animal activity trajectory data to generate a standardized activity trajectory sequence; Performing time alignment and outlier correction on the habitat environmental parameter sequence to generate a continuous environmental parameter surface; The step of dynamically matching the animal activity trajectory data with a baseline threat pattern in a predefined ecological threat indicator library to generate a first ecological association network feature includes: Performing spatiotemporal grid mapping on the standardized activity trajectory sequence and the continuous environmental parameter surface to extract the behavior density distribution and environmental parameter gradient within the grid unit; The behavior density distribution and the environmental parameter gradient are compared with the spatiotemporal constraints of the baseline threat pattern by a threat pattern matching engine to generate a threat matching confidence score for each grid cell; A three-dimensional threat heat distribution map is constructed based on the threat matching confidence of all grid cells as the first ecological association network feature.

3. The method for monitoring and early warning of wildlife nature reserves according to claim 2, characterized in that: The step of extracting migration path segments from the animal activity trajectory data comprises: Identifying path turning points and dwell periods in the standardized activity trajectory sequence, and segmenting migration path subsegments with continuous motion vectors; Perform curvature analysis and speed variation detection on each migration path sub-segment, and mark abnormal path morphological features; The step of generating the second ecological association network feature includes: Perform causal correlation analysis on the abnormal path morphological characteristics and the environmental parameter mutation events within the corresponding time window, and calculate the path-environment coupling coefficient; Based on the path-environment coupling coefficients of all migration path subsegments, a dynamic threat transmission chain was constructed as the second ecological association network feature; Furthermore, the step of performing cross-modal feature fusion on the first ecological association network feature and the second ecological association network feature to generate a multi-level threat coupling vector includes: performing principal component dimensionality reduction processing on the three-dimensional threat heat distribution map to extract core threat space vectors; The dynamic threat propagation chain is converted into a spatiotemporal propagation probability matrix, and a tensor product operation is performed with the core threat space vector to generate the multi-level threat coupling vector.

4. The method for monitoring and early warning of wildlife nature reserves according to claim 1, characterized in that: The step of constructing a voiceprint anomaly detection map based on the frequency domain feature distribution of the biometric voiceprint signal set includes: Performing frame and window processing on the biometric voiceprint signal set to extract the Mel-frequency cepstral coefficients and harmonic energy ratios of each frame signal; The abnormal voiceprint classifier is used to jointly identify the Mel-frequency cepstral coefficient and the harmonic energy ratio, and output a voiceprint abnormality probability distribution map; Based on the sound source localization algorithm, abnormal voiceprint events are spatially clustered to generate a voiceprint anomaly detection map with orientation marks; Furthermore, the step of performing spatial superposition analysis on the voiceprint anomaly detection map and the multi-level threat coupling vector to generate a comprehensive threat situation matrix includes: Registering the orientation mark of the voiceprint anomaly detection map with the spatial coordinates of the multi-level threat coupling vector; The overlapping area ratio and spatial correlation index of the voiceprint abnormal area and the threat coupling area are calculated to generate the comprehensive threat situation matrix.

5. The method for monitoring and early warning of wildlife nature reserves according to claim 1, characterized in that: The step of dynamically calibrating the threshold of the comprehensive threat situation matrix using an adaptive weight distribution algorithm and outputting a real-time threat level spectrum includes: Based on the threat level distribution patterns in the historical threat event database, a baseline threshold curve for each threat dimension is established; The sliding time window algorithm is used to calculate the cumulative intensity value of each threat dimension in the current comprehensive threat situation matrix in real time; Mapping the cumulative intensity value onto the baseline threshold curve by a nonlinear interpolation method to output a dynamically adjusted real-time threat level spectrum; Furthermore, according to the mapping relationship between the real-time threat level spectrum and the preset emergency response strategy library, activating the corresponding level of ecological protection intervention instruction set includes: Analyze the weight distribution of each threat dimension in the real-time threat level spectrum and match the multi-level response protocol in the preset emergency response strategy library; Generate drone patrol path planning, ecological corridor blocking instructions, and manual inspection priority queues based on the spatial coverage and duration of the threat level; The instruction set is synchronized to the protected area management terminal and mobile law enforcement equipment via a low-latency communication network.

6. The method for monitoring and early warning of wildlife nature reserves according to any one of claims 1 to 5, characterized in that: The method further comprises: Establishing a digital twin model of the ecosystem of the target wildlife nature reserve, and receiving and mapping the multimodal ecological monitoring data stream in real time; A threat evolution simulator is embedded in the digital twin model to predict the threat diffusion path within the future time window based on the current comprehensive threat situation matrix; Optimizing the execution sequence and resource allocation plan of the ecological protection intervention instruction set based on the prediction results; The updating step of the digital twin model includes: Collect ecological status feedback data after the actual intervention measures are implemented and calculate the threat suppression efficiency coefficient; Inputting the threat suppression efficiency coefficient into the threat evolution simulator for back propagation training to update the prediction parameters of the threat diffusion path; When the threat suppression efficiency coefficient is lower than a preset threshold, the secondary response mechanism is triggered and the cross-regional collaborative intervention instruction set is regenerated.

7. The method for monitoring and early warning of wildlife nature reserves according to claim 1, characterized in that: The method further comprises: Establish a human activity monitoring network around the target wildlife nature reserve to collect traffic flow data, nighttime light intensity distribution, and illegal intrusion alarm signals; Cross-validate human activity monitoring data with the comprehensive threat situation matrix to identify complex ecological threats induced by human factors; Automatically generate law enforcement evidence collection clue packages and public warning information push strategies based on complex threat characteristics; The step of cross-validating the human activity monitoring data with the comprehensive threat situation matrix to identify complex ecological threats induced by human factors includes: Analyze the spatiotemporal coupling between the spatiotemporal trajectories of human activities and abnormal animal behavior events, and calculate the contribution factor of human interference; Assess the long-term cumulative effects of human activities on ecological threats based on persistent trends in habitat environmental parameters; When the human interference contribution factor exceeds a critical value, the directional tracking and image acquisition functions of the high-precision video surveillance equipment are activated.

8. The method for monitoring and early warning of wildlife nature reserves according to claim 1, characterized in that: The method further comprises: Deploy edge computing devices at key ecological nodes in the target wildlife nature reserves to perform real-time feature extraction on locally collected multimodal ecological monitoring data streams; The feature extraction results of each node are encrypted and aggregated through the federated learning framework to update the parameters of the global threat detection model; Dynamically distribute updated model parameters to each edge computing device to achieve the collaborative evolution of distributed threat perception capabilities; The updating steps of the federated learning framework include: Design a differential privacy protection strategy to add geo-masking noise to sensitive animal location information; Adopt model parameter differential aggregation technology to eliminate the impact of edge node data distribution skew on the global model; When a node device is detected to be abnormally offline, the model parameter rollback mechanism is activated and the computing tasks are reallocated to adjacent nodes.

9. The method for monitoring and early warning of wildlife nature reserves according to claim 1, characterized in that: The method further comprises: Integrate historical ecological restoration records and species reproduction data of the target wildlife nature reserves to construct an ecosystem resilience assessment matrix; Conduct multidimensional correlation analysis between the real-time threat level spectrum and the ecosystem resilience assessment matrix to predict the ecological recovery potential under different intervention strategies; Dynamically adjust the implementation intensity and effect period of the ecological protection intervention instruction set according to the prediction results; The step of performing a multidimensional correlation analysis between the real-time threat level spectrum and the ecosystem resilience assessment matrix to predict the ecological restoration potential under different intervention strategies includes: A threat intensity-restoration efficiency response surface model was developed to quantify the restoration rates of key ecosystem indicators under different intervention measures. Simulate the transition path of ecological resilience under the superposition of multiple rounds of intervention measures to identify the optimal combination of intervention strategies; When the restoration rate is predicted to be lower than the degradation rate, an expert consultation request is triggered and a protected area zoning management plan is initiated.

10. A monitoring and early warning system for wildlife nature reserves, characterized in that: The wildlife nature reserve monitoring and early warning system includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the wildlife nature reserve monitoring and early warning method described in any one of claims 1 to 9 above.

Images