A method and system for predicting a mountain collapse

Abstract

The application provides a mountain collapse prediction method and a collapse prediction system, and relates to the technical field of mountain collapse prediction. The method comprises the following steps: first, obtaining animal data of a target mountain area, screening out target resident animals, determining monitoring points according to the daily activity range of the target resident animals, and implementing monitoring. Then, collecting historical and real-time daily activity video data of the target resident animals, extracting behavior characteristics, comparing and obtaining distinguishing characteristics, and matching the distinguishing characteristics to obtain a matching score. When the matching score exceeds a set threshold, a geological sensor is controlled to collect geological data. If the data is abnormal, it is determined that a mountain collapse will occur, and warning information is generated in combination with the distinguishing characteristics. Early warning is realized by using the sensitivity of animals to geological changes, and accurate judgment is realized through geological data.

Description

Technical Field

[0001] This application relates to the field of landslide prediction technology, and in particular to a landslide prediction method and landslide prediction system. Background Technology

[0002] In the face of frequent geological disasters, landslides pose a serious threat to human life, property, and the ecological environment, making efficient and accurate prediction technology crucial for disaster prevention and mitigation.

[0003] Currently, the most common existing technology for predicting landslides relies on deploying geological sensors in mountainous areas to monitor physical parameters such as displacement, stress, and crack width in real time. By setting thresholds, a landslide warning is issued when the monitored data exceeds the threshold.

[0004] However, this existing technology has significant technical problems. Changes in monitoring data usually occur in the immediate stage of a landslide, and the early warning window is too short to meet the needs of early evacuation and prevention. This makes it difficult to effectively implement disaster prevention and mitigation measures and cannot fundamentally reduce the losses caused by landslides. Summary of the Invention

[0005] This application provides a landslide prediction method and landslide prediction system, which can accurately identify potential landslide risk areas and obtain key geological data in a timely manner by comprehensively analyzing changes in the behavior of resident animals and the geological structure characteristics of mountains, thereby achieving early warning of landslides.

[0006] In a first aspect, this application provides a method for predicting landslides. The method includes: acquiring animal data for a target mountain area, filtering the animal data to identify multiple different target resident animals, the animal data including at least the animals' daily activity range; determining monitoring points based on the daily activity range of each target resident animal, and monitoring each target resident animal based on the monitoring points; acquiring historical daily activity video data of each target resident animal within a set time period, and extracting corresponding historical behavioral characteristics of each target resident animal; acquiring real-time daily activity video data of each target resident animal, and extracting corresponding real-time behavioral characteristics of each target resident animal; comparing the real-time behavioral characteristics of each target resident animal with the historical behavioral characteristics to determine distinguishing features; matching the distinguishing features of each target resident animal to obtain a matching score; if the matching score is greater than a set matching score threshold, controlling multiple pre-set geological sensors to collect geological data; if the geological data shows an anomaly, determining that a landslide is imminent, and generating an abnormal animal behavior warning based on the distinguishing features.

[0007] By adopting the above technical solution, animal data for the target mountain area is first acquired, and resident animals are screened. Monitoring points are determined based on the animals' daily activity range. By comparing historical and real-time behavioral characteristics, abnormal animal behavior can be detected in advance. When the matching score exceeds a threshold, geological sensors are activated, and early warning information is generated when an anomaly occurs. This process constructs a dual monitoring system of "initial screening of animal behavior - verification of geological data." It utilizes the sensitivity of animals to geological changes to achieve early warning, and then uses geological data for precise judgment. Compared with a single monitoring method, this significantly improves the timeliness and accuracy of landslide prediction and effectively reduces the risk of disaster.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, the acquisition of animal data of the target mountain area and the screening of the animal data to determine multiple different target resident animals include: acquiring the infrasound perception sensitivity of each animal in the animal data and the activity range curve of each animal; calculating the perception weight of the infrasound perception sensitivity and the habitat stability weight of the activity range curve; and combining the perception weight and the stability weight to select the top N animals as target resident animals.

[0009] By employing the above technical solution, when selecting target resident animals, a combination of infrasound sensing sensitivity and activity range curves is used to calculate sensing weights and stability weights for animal selection. Animals with high infrasound sensing sensitivity can detect infrasound generated by geological activity earlier, while animals with stable activity ranges can avoid monitoring interference caused by migration and other factors. Combining these two weighting methods ensures that the selected animals are both sensitive to geological changes and can provide stable monitoring data over a long period, thereby improving the reliability and effectiveness of monitoring data from the source and laying a solid foundation for accurate prediction of landslides through animal behavior.

[0010] In some embodiments of the first aspect, matching the distinguishing features of each target resident animal to obtain a matching score includes: acquiring the time series and spatial series of each distinguishing feature; performing the following operations on any two animals among the target resident animals: calculating the time interval between the first target resident animal and the second target resident animal based on the time series, and constructing a first activity trajectory of the first target resident animal and a second activity trajectory of the second target resident animal based on the spatial series; comparing the time interval with a preset time window threshold to determine a time synchronization score; calculating the overlap area and spatial position deviation of the first activity trajectory and the second activity trajectory to determine a spatial consistency score; calculating the behavioral correlation score between the first target resident animal and the second target resident animal based on the time synchronization score and the spatial consistency score; and integrating the behavioral correlation scores between all pairs of animals among the target resident animals to obtain the matching score.

[0011] By employing the above technical solution, the time and spatial sequences of distinguishing features are obtained when calculating the matching score. Pairwise analysis of temporal synchronicity and spatial consistency scores for each animal is performed, and these scores are integrated to obtain the matching score. The temporal synchronicity score determines whether abnormal animal behaviors occur simultaneously, while the spatial consistency score measures the degree of correlation between activity trajectories. By comprehensively assessing the behavioral correlation using both methods, accidental behavior of individual animals can be effectively ruled out. Through cross-validation of multiple animal behaviors, it is possible to accurately determine whether abnormal animal behavior is caused by geological activity, avoiding misjudgments and greatly improving the accuracy and reliability of landslide early warning systems.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, before the step of controlling multiple pre-set geological sensors to collect geological data, the method further includes: acquiring mountain geological structure data of the mountain area; performing density clustering analysis on the spatial sequence to construct an aggregation area of ​​abnormal animal behavior; determining the area of ​​abnormal animal behavior based on the spatial boundary of the aggregation area; performing spatial overlay analysis on the area of ​​abnormal animal behavior and the mountain geological structure data to determine the geological structural weak zone corresponding to the area of ​​abnormal animal behavior; and determining the location information of the landslide risk area based on the geological structural weak zone.

[0013] By employing the above technical solution, geological structure data of the mountain is acquired before activating the geological sensors. Spatial sequence clustering analysis of abnormal animal behavior identifies clustering areas and anomalous regions. Overlay analysis with geological structure data identifies weak zones and areas at risk of landslides. Geological structure data provides inherent risk information about the mountain, while abnormal animal behavior reflects real-time changes. Combining the two allows for precise location of areas with weak geological structures and abnormal animal behavior, narrowing the monitoring range and making subsequent geological sensor monitoring more targeted, thereby improving monitoring efficiency and prediction accuracy.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, after the step of determining the location information of the collapse risk area based on the weak geological structure zone, the method further includes: obtaining the two target resident animals with the highest scores in each behavior correlation score; obtaining the activity trajectories corresponding to the two target resident animals, and determining the danger source area based on the activity trajectories; determining the overlapping area between the location information and the danger source area; and controlling the geological sensor located in the overlapping area to collect geological data.

[0015] By employing the above technical solution, after determining the location information of the landslide risk area, the two animals with the highest behavioral correlation scores are selected. Their activity trajectories are used to identify the hazard source area and find the overlapping area with the location information. Data is then collected by sensors in this area. Animals with high scores exhibit more representative behavior; tracing their activity trajectories back to the identified hazard source area points to the potential landslide initiation point. When this overlaps with the risk area, key monitoring areas are precisely located, enabling geological sensors to focus on key locations for data collection, reducing invalid monitoring, and improving data effectiveness and prediction efficiency.

[0016] In some embodiments of the first aspect, determining the hazard source area based on the activity trajectory includes: performing time synchronization and spatial coordinate calibration on the activity trajectory data to generate a three-dimensional motion path with a timestamp; tracing back the trajectory in the opposite direction of the motion path to construct a set of reverse motion vectors; calculating the convergence region of each reverse motion vector in three-dimensional space, the convergence region being the region where the spatial distribution density of the intersection points of the reverse vectors exceeds a preset distribution density threshold; and determining the convergence region as the hazard source area.

[0017] By employing the above technical solution, when identifying the hazard source area, the activity trajectory is synchronized in time and calibrated spatially. After generating a three-dimensional path, a vector set is constructed through reverse backtracking, and the convergence region is calculated. Time synchronization and spatial calibration ensure the accuracy of the trajectory data, and reverse backtracking simulates the animal's escape direction. The convergence region represents the source of the animal's escape, reflecting the potential initiation location of the collapse. This method accurately locates the hazard source based on animal behavior logic, providing accurate evidence for subsequent targeted monitoring and early warning, and enhancing the reliability of predictions.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, if the geological data shows an anomaly, it is determined that a landslide is about to occur, including: obtaining a geological data feature set of historical landslides; matching the geological data with the geological data feature set to obtain a landslide matching score; if the landslide matching score is greater than a set mountaintop landslide matching score threshold, it is determined that the geological data shows an anomaly.

[0019] By employing the aforementioned technical solution, when determining whether geological data is abnormal, real-time data is matched with a set of historical landslide geological data feature sets to obtain a matching score. If the score exceeds a threshold, an anomaly is confirmed. The historical feature set has accumulated a large amount of landslide data patterns. Matching new data with this set can quickly determine whether the current geological conditions conform to landslide characteristic patterns, avoiding subjective judgment. Through quantitative data analysis, the objectivity and accuracy of geological data anomaly judgment are improved, thereby ensuring the scientific validity and credibility of landslide prediction results.

[0020] In a second aspect, this application provides a landslide prediction system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the landslide prediction system to perform the methods described in the first aspect and any possible implementation thereof.

[0021] Thirdly, this application provides a computer-readable storage medium including instructions that, when executed on a collapse prediction system, cause the collapse prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, this application provides a computer program product that, when run on a collapse prediction system, causes the collapse prediction system to perform the method described in the first aspect and any possible implementation thereof.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] 1. By employing a technical approach that selects target resident animals based on their daily activity range, compares real-time behavioral characteristics with historical characteristics, matches distinguishing features of multiple animals, and verifies the results in conjunction with geological sensors, the technical problems of excessively short warning time windows and high monitoring costs in existing technologies are effectively solved. This enables early and accurate warnings of landslides, reducing monitoring costs and disaster risks.

[0025] 2. By employing a technique based on time and spatial sequences with distinctive features to calculate temporal synchronicity scores and spatial consistency scores among animals and integrating them into a matching score, this technique effectively solves the technical problem in existing technologies where it is difficult to distinguish whether abnormal animal behavior is caused by geological activity or accidental factors, leading to a high rate of false alarms. This allows for accurate judgment of anomalies caused by geological activity through cross-validation of multiple animal behaviors, significantly improving the reliability and accuracy of landslide early warnings.

[0026] 3. By employing a technique that uses behavioral correlation scores to select representative animals, traces their activity trajectories to determine the source of danger, and overlaps with the landslide risk area, and precisely controls the data collected by sensors, this technique effectively solves the technical problems of existing geological sensors, such as large monitoring range, low data validity, and difficulty in locating key landslide areas. This allows for focusing on key monitoring areas, reducing ineffective monitoring, and significantly improving data validity and prediction efficiency. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a landslide prediction method in an embodiment of this application;

[0028] Figure 2 This is a schematic diagram of an application scenario of the landslide prediction method in the embodiments of this application;

[0029] Figure 3 This is a schematic diagram of an application scenario of the landslide prediction method in the embodiments of this application;

[0030] Figure 4 This is another flowchart illustrating the landslide prediction method in the embodiments of this application;

[0031] Figure 5 This is a schematic diagram of the physical device structure of a collapse prediction system in the embodiments of this application. Detailed Implementation

[0032] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0033] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0034] For ease of understanding, the method provided in this implementation is described in process below. Please refer to [link / reference]. Figure 1 This is a flowchart illustrating a landslide prediction method in an embodiment of this application.

[0035] S101. Obtain animal data for the target mountain area, and screen the animal data to identify multiple different target resident animals. The animal data shall at least include the animals' daily activity range.

[0036] The term "target mountain area" refers to the specific mountain range selected as the subject of landslide prediction research. "Animal data" refers to a collection of information on animals within the target mountain area, covering their species, numbers, habits, physiological characteristics, and daily activity range. "Target resident animals" refers to a variety of different animals that, after screening, are identified as having long-term stable survival and activity within the target mountain area; these will serve as important biological indicators for monitoring landslides. "Daily activity range" represents the geographical area within which animals normally engage in activities such as foraging, drinking, resting, and socializing. For example, a wild boar forages and moves about in a forest area within a radius of several kilometers; this area constitutes its daily activity range.

[0037] Specifically, when initiating landslide prediction work, the primary task of the landslide prediction system is to obtain animal data for the target mountain area. The system can collect this data through multiple methods. Firstly, it deploys professional field survey teams equipped with telescopes, infrared cameras, GPS trackers, and other equipment to penetrate deep into the target mountain area. Surveyors observe animals over extended periods, recording their species and behaviors using telescopes; capturing images of nocturnal animals using infrared cameras; and equipping some animals with GPS trackers to monitor their movements in real time, thereby obtaining information such as their daily activity range. Secondly, the system integrates existing data resources, including animal census records from local forestry departments, animal research findings from universities or research institutions in the area, and monitoring data from wildlife conservation organizations. After collecting animal data, the system enters a screening phase to identify target resident animals. During the screening process, the system not only considers the animals' daily activity range to ensure their activities are mainly concentrated within the target mountain area but also assesses their ability to perceive environmental changes. Many animals are highly sensitive to warning signs of landslides, such as infrasound and subtle vibrations. The system retrieves data on the sensitivity of different animals to infrasound by consulting animal physiological research. Simultaneously, it analyzes animal activity patterns to plot activity range curves, thereby assessing the stability of animal habitats. For example, migratory birds with wide ranges that frequently leave the target mountain area are unsuitable as permanent resident animals due to their behavior being highly susceptible to external disturbances. Animals like the Himalayan blue sheep, which live long-term on the steep cliffs of the target mountain and are sensitive to environmental changes, are given priority. By integrating various indicators and employing scientific algorithms and evaluation models, the system ultimately selects several different target permanent resident animals from a large pool of potential species. These animals will subsequently become important biological indicators for monitoring landslides.

[0038] S102. Based on the daily activity range of each resident animal of the target, determine the monitoring point and monitor each resident animal of the target based on the monitoring point.

[0039] "Monitoring points" refer to specific locations within the daily activity range of the target resident animal, determined through scientific planning and precise site selection, where various monitoring devices are installed to collect animal behavioral data. Examples include monitoring points set up in the center of areas where animals frequently forage, near nests, and next to watering points. "Monitoring" refers to the continuous or periodic observation, recording, and data collection of the target resident animal's behavioral dynamics, activity trajectories, and sound changes using various technologies such as high-definition cameras, infrared sensors, sound collectors, and radar monitoring equipment, thereby obtaining the animal's behavioral characteristics at different times and in different scenarios.

[0040] Specifically, once the landslide prediction system identifies the target resident animals and their daily activity range, it begins determining the monitoring locations. This process requires the system to comprehensively consider multiple factors to ensure the accurate and complete capture of the target resident animals' behavioral information. First, the system conducts an in-depth analysis of the activity habits and behavioral patterns of each target resident animal. For animals with relatively fixed and concentrated activity ranges, such as sheep flocks foraging and resting in fixed areas, the system sets up monitoring points in key locations such as the sheep's main foraging grasslands and resting slopes to ensure clear recording of the sheep's feeding, movement, and resting behaviors. For animals with wide activity ranges and complex activity patterns, such as squirrels traversing entire forests, the system employs a combination of zoned and focused monitoring. The squirrel's activity range is first divided into multiple sub-regions. Within each sub-region, monitoring points are set up in key locations such as densely wooded areas and nut-growing areas where squirrels frequently operate. Monitoring points are also set up at connecting passages between different sub-regions, forming a comprehensive monitoring network. When determining monitoring locations, the system also fully considers the impact of topography and environmental conditions on the monitoring effectiveness. In areas with complex terrain and obstructed visibility, high-definition cameras with wide-angle shooting capabilities are selected and installed at appropriate angles to expand the monitoring range. In areas with insufficient light at night, infrared night vision cameras and infrared sensors are used to ensure effective monitoring of animal activity even in dark environments. For animals with rich sound signals, such as birds, sound collectors are placed in their activity areas to record their calls, flight sounds, and other auditory information. After the monitoring points are determined, the system installs the corresponding monitoring equipment at these points and conducts rigorous debugging and calibration. For example, the focal length, shooting angle, and image clarity of the cameras are adjusted; the sensing range and sensitivity of the infrared sensors are tested; and the volume and sound quality of the sound collectors are optimized to ensure that all monitoring equipment can operate stably and accurately. After the equipment is installed and debugged, the system continuously monitors the resident animals at these monitoring points, collecting animal behavior data at set time intervals or in real time. The acquired video, image, and sound data are transmitted to the data processing center, providing detailed basic data for subsequent analysis of the relationship between animal behavior and landslides.

[0041] In some embodiments, in scenarios where real-time monitoring of the monitoring area is conducted routinely, the monitoring equipment may not be continuously powered on. Instead, it can be activated only when abnormal animal activity is confirmed, thus saving resources. Specifically, the system uses multiple infrared sensors pre-positioned at monitoring points to detect the presence of animal heat source signals in the surrounding preset area. The infrared sensors continuously scan the surrounding environment, and when they detect infrared radiation matching the characteristics of an animal heat source, they trigger the subsequent sound acquisition process. When an animal heat source signal is detected, the system controls the sound sensors to collect target sound data. The sound sensors begin working upon triggering the infrared sensors' signal detection, collecting sounds from the surrounding environment. The collected target sound data is transmitted to the system's analysis module and compared with the sound frequency characteristics of the target resident animal. If the target sound data matches the sound frequency characteristics of the target resident animal, it indicates that the target resident animal may be present in the monitoring area. At this time, the system triggers the monitoring equipment corresponding to that monitoring point to collect video data of the target resident animal. After the monitoring equipment is activated, the system identifies the collected video data within a set startup time to determine whether the animal corresponding to the animal heat source signal is the target resident animal. This identification process utilizes image recognition technology to analyze and judge the animal's appearance and behavioral characteristics in the video. If the identification result is positive, confirming it as the target resident animal, the system will continue to control the monitoring equipment to collect video data for further observation of its behavioral changes. If the identification result is negative, indicating that the animal corresponding to the detected heat source signal is not the target resident animal, the system will control the monitoring equipment to stop working to save resources. Furthermore, when the target resident animal leaves the set monitoring range, or when no animal heat source signal is detected within a preset waiting time, the system will also control the monitoring equipment to stop working. This is because in these situations, continuing to monitor the area is unnecessary, and stopping the equipment can effectively reduce the system's energy consumption and resource usage. Specifically, an infrared sensor is a device that can sense infrared radiation and convert it into an electrical signal, used to detect changes in heat in the surrounding environment, thereby discovering potential animal heat source signals; an animal heat source signal refers to a specific infrared radiation pattern formed by the heat generated by the animal's body; a sound sensor is a device used to collect sound signals and convert them into electrical signals, which can be pre-positioned near the monitoring point. Implementing the above measures will not only enable timely acquisition of data on abnormal animal behavior, providing a basis for landslide prediction, but also avoid resource waste, significantly improve the efficiency and accuracy of system monitoring, and build a solid data collection defense line for geological disaster early warning.

[0042] S103. Obtain historical daily activity video data of each target resident animal within a set time period, and extract the corresponding historical behavioral features of each target resident animal; obtain real-time daily activity video data of each target resident animal, and extract the corresponding real-time behavioral features of each target resident animal.

[0043] "Set time" refers to a pre-determined time period used to define the time range for collecting historical daily activity video data. This time period can be flexibly set according to actual monitoring needs, such as one month, six months, or one year. Its purpose is to obtain data on the routine behavioral patterns of the target resident animal within a relatively stable period. "Historical daily activity video data" refers to video information of the animal's daily activities recorded within the set time period by equipment (such as high-definition cameras, infrared cameras, etc.) deployed at monitoring points within the target resident animal's daily activity range. These videos record in detail the animal's activities at different times and in different scenarios, such as video clips of wild boars foraging, resting, and moving in the forest every day over the past six months. "Historical behavioral characteristics" refers to a series of characteristic information extracted from the historical daily activity video data using specific analysis techniques and algorithms, reflecting the target resident animal's routine behavioral patterns within the set time period. This includes the animal's activity patterns (such as activity time and frequency), movement trajectory, behavior type (foraging, resting, socializing, etc.), and behavior duration. "Real-time daily activity video data" refers to video information of the daily activities of the target resident animal collected in real time by the monitoring equipment, such as real-time footage of a squirrel jumping and foraging in the forest. "Real-time behavioral characteristics" refers to the feature information extracted from the real-time daily activity video data that reflects the current behavioral state of the target resident animal. This feature information reflects the animal's current behavioral performance.

[0044] Specifically, once the landslide prediction system has completed the setup of monitoring points and ensured the equipment is operating normally, it begins to continuously collect video data on the daily activities of the target resident animals. For historical daily activity video data, the system periodically organizes and filters the video data stored in the database according to a pre-set time period. For example, if the time period is set to three months, the system will automatically summarize and categorize all video data of the target resident animals captured at each monitoring point within that period at the end of each three-month period. Then, the system uses advanced image recognition, computer vision, and deep learning algorithms to conduct in-depth analysis of this historical video data. Taking deer herds as an example, the system analyzes the movement trajectories of the deer in the videos and uses trajectory tracking algorithms to determine their daily activity routes; by statistically analyzing the duration and frequency of behaviors such as eating, drinking, and resting, it constructs the behavioral pattern characteristics of the deer herd within that time period. For real-time daily activity video data, the system uses real-time video streaming technology to transmit the video data captured by the monitoring equipment to the data processing center in real time with minimal latency. Once the real-time video data is received, the system immediately initiates the real-time behavioral feature extraction program. The system, based on deep learning models such as convolutional neural networks (CNNs), performs real-time identification and analysis of animal behavior in videos. For example, when birds are detected, the system can quickly identify their flight posture and foraging movements, and extract real-time behavioral characteristics such as current flight speed, foraging frequency, and flight altitude. The system continuously acquires and extracts historical and real-time behavioral features, providing dynamic and comprehensive data support for subsequent analysis of changes in animal behavior. This enables the timely detection of abnormal animal behavior and provides crucial information for landslide prediction.

[0045] S104. Compare the real-time behavioral characteristics of each resident animal of the target with the historical behavioral characteristics to determine the distinguishing features;

[0046] Among them, "distinguishing features" are used to represent feature information that differs between real-time behavioral features and historical behavioral features. These differences may be reflected in multiple dimensions such as the temporal distribution of behavior, behavioral patterns, activity range, and behavioral frequency. For example, an animal that originally foraged in a fixed area may suddenly expand its activity range, or the animal's activity time may change significantly. These changes are all distinguishing features.

[0047] Specifically, after acquiring the real-time and historical behavioral characteristics of the target resident animals, the collapse prediction system conducts a detailed comparative analysis of each behavioral characteristic. Regarding activity time characteristics, the system compares the animal's historical daily start and end times with the currently monitored activity times. For example, if a monkey typically starts its activity around 6 AM in the past six months, but recently delays its activity until 8 AM for several consecutive days, this significant change in activity time will be marked as a distinguishing feature. For activity range characteristics, the system compares the area covered by the animal's historical activity trajectory with its real-time activity trajectory. For instance, if a wolf's activity range was previously concentrated in a valley, but real-time monitoring shows it frequently appearing on distant hillsides surrounding the valley, this expansion of its activity range will also be considered a distinguishing feature. Regarding behavioral patterns, the system analyzes whether there have been changes in the animal's movement patterns, social behaviors, etc. For example, if a herd of antelopes that previously moved in groups is now observed to have some individuals acting alone, or if the antelopes' running posture or jumping frequency shows abnormalities, these changes in behavioral patterns will also be recorded as distinguishing features. During the comparison process, the system employs statistical methods and machine learning algorithms to quantitatively assess the degree of difference in various features. For subtle changes, the system combines multiple relevant features for comprehensive judgment, avoiding misjudgments due to minor fluctuations in a single feature. By comprehensively and deeply comparing real-time and historical behavioral features, the system can accurately identify distinguishing features with practical significance. These distinguishing features will become important clues for subsequent assessments of whether a mountain faces a risk of collapse, providing crucial evidence for further analysis and decision-making.

[0048] S105. Match the distinctive features of each of the target resident animals to obtain a matching score;

[0049] This step is performed after the system has determined the distinguishing characteristics of each target resident animal. It is executed to further assess the synergy and correlation of abnormal animal behavior, thereby determining the likelihood of landslide risk. The scenario is within the data processing and analysis phase of the landslide prediction system, where the system comprehensively analyzes the distinguishing characteristics of animal behavior. This step will be explained in detail in subsequent steps S401-S406 and will not be repeated here.

[0050] S106. If the matching score is greater than the set matching score threshold, control multiple pre-set geological sensors to collect geological data.

[0051] The "matching score" refers to a quantitative score obtained by matching the distinguishing features of each target resident animal. It reflects the degree of correlation and synergy between abnormal behaviors of different target resident animals. The "matching score threshold" represents a pre-set reference score value used as a standard to determine whether to activate the geological sensor to collect geological data. This threshold is determined based on a comprehensive analysis of extensive experimental data, historical cases, and expert experience. Optionally, the matching score threshold can be determined as follows: Collect past landslide events in the target mountain area and similar geological environments, extracting matching scores for abnormal animal behaviors before the events; simultaneously collect matching scores for abnormal animal behaviors (such as occasional individual behaviors) that did not occur during landslides, forming matching score datasets for "landslide-non-landslide" samples. By plotting ROC curves, calculate the true positive rate (correctly identifying abnormal behaviors before landslides) and false positive rate (mistakenly identifying occasional behaviors as landslide precursors) at different thresholds, and select the score that results in the highest true positive rate and lowest false positive rate as the initial threshold. For example, if statistics show that 90% of landslide events have a matching score above 80, while 95% of non-landslide cases have a matching score below 80, then the initial threshold can be tentatively set at 80. Then, geological disaster experts and animal behavior experts are invited to adjust the initial threshold based on regional geological characteristics (such as rock stability and fault distribution) and animal habits (such as sensory sensitivity and activity patterns). For example, for areas with extremely unstable geological structures, the threshold can be appropriately lowered to improve early warning sensitivity; for areas where animal behavior is easily affected by external factors (such as frequent tourism activities), the threshold can be appropriately increased to reduce false alarms.

[0052] Specifically, after calculating the matching score, the landslide prediction system compares it with a pre-set matching score threshold. If the matching score is less than or equal to the threshold, it indicates a weak correlation between the behavioral abnormalities of the target resident animals, possibly due to accidental factors or individual differences. In this case, the system determines that the current landslide risk is low and will not activate geological sensors to collect geological data, but will continue to monitor animal behavior, awaiting the next round of data collection and analysis. However, if the matching score is greater than the threshold, it indicates a high degree of synergy and correlation in the behavioral abnormalities of multiple target resident animals in time and space, likely due to changes within the mountain (such as stress changes, rock loosening, etc.) causing the animals to perceive the abnormalities and exhibit behavioral changes. In this case, the system determines that the mountain has a high risk of landslide. In this situation, the system immediately issues a command to control multiple geological sensors pre-set within the target mountain area to begin collecting geological data. After receiving the system command, these geological sensors will collect various geological parameters of the mountain in real time according to the preset sampling frequency and data transmission method. For example, displacement sensors monitor minute displacement changes on the mountain surface, pressure sensors detect stress changes inside the mountain, vibration sensors capture vibration signals from the mountain, and humidity sensors measure soil moisture. The collected data is transmitted in real time to a data processing center via wireless or wired communication networks so that the system can further analyze this geological data to determine whether the mountain is truly about to collapse, thus providing accurate geological evidence for subsequent early warning and decision-making.

[0053] In some embodiments, after obtaining a matching score for the behavior of a target resident animal that is greater than a set matching score threshold, the scenario involves further precise location of the landslide risk area when a potential landslide risk is identified in a mountain. First, the system acquires geological structure data of the mountain area from sources such as geological exploration data, satellite remote sensing data, and Geographic Information Systems (GIS). This data may include detailed information such as rock hardness, rock strata tilt angle, groundwater distribution, and fault zone location. Simultaneously, the system retrieves previously acquired spatial sequence data of abnormal animal behavior. Next, a density clustering analysis algorithm is used to process the spatial sequence. Taking the DBSCAN (Density-Based Spatial Clustering) algorithm as an example, it divides densely connected points in space into a cluster. When processing the spatial sequence of abnormal animal behavior, the algorithm scans all spatial location points. When the number of abnormal animal behavior location points in a certain area reaches or exceeds a set density threshold, and the distance between these points is within a set neighborhood radius, these points are grouped into an abnormal animal behavior cluster. For example, in a certain forest area, multiple target resident animals exhibit abnormal running and gathering behaviors within a short period of time. After density clustering analysis, the spatial locations of these behaviors form a cluster area.

[0054] Then, based on the spatial boundaries of the clustered areas, the system expands outward or integrates adjacent clustered areas to determine a larger area of ​​abnormal animal behavior, encompassing all spatial ranges where abnormal animal behavior is clearly observed. Next, the system performs spatial overlay analysis of the abnormal animal behavior area and mountain geological structure data on a GIS platform. By comparing the spatial distribution of the two, it identifies overlapping areas between the abnormal animal behavior area and weak points in the mountain's geological structure; these overlapping areas are designated as geologically weak zones. For example, if the abnormal animal behavior area overlaps with fault areas or severely weathered rock areas within the mountain, these overlapping areas are identified as geologically weak zones. Finally, based on the identified geologically weak zones, the system combines information such as the mountain's topography and geomorphology to further clarify the location information of landslide risk areas, specifying precise coordinate ranges or geographical descriptions, providing accurate data for subsequent targeted monitoring and early warning. Throughout this process, the accuracy of the data and the rationality of the algorithm are crucial. For mountain geological structure data, regular updates and verification are necessary to ensure it accurately reflects the mountain's geological conditions. For density clustering analysis algorithms, parameters such as density thresholds and neighborhood radii need to be set appropriately based on actual conditions to ensure accurate identification of areas where abnormal animal behavior occurs. Furthermore, when performing spatial overlay analysis, coordinate system transformation and accuracy matching issues for different data sets must be considered to ensure the reliability of the analysis results.

[0055] In some embodiments, after determining the location information of the landslide risk area, the system selects the pair of animals with the highest pairwise behavioral correlation scores from all target resident animals. For example, assuming there are 5 target resident animals, and the pairwise behavioral correlation scores are: goat-squirrel 90 points, goat-hare 85 points, squirrel-hare 78 points, etc., then goat and squirrel are the two animals with the highest scores. Next, the system retrieves the activity trajectory data corresponding to these two animals. These trajectories are constructed based on their spatial sequences and may contain three-dimensional coordinate points corresponding to multiple timestamps. Then, the danger source area is determined based on the activity trajectory. The specific process may include: synchronizing the activity trajectory data in time and calibrating its spatial coordinates to generate a timestamped 3D movement path, ensuring the temporal and spatial accuracy of the trajectory; tracing back the trajectory in the opposite direction to construct a set of reverse movement vectors. For example, if an animal moves from point A to point B and then to point C, the reverse tracing would be from point C to point B and then back to point A; calculating the convergence region of each reverse movement vector in 3D space, i.e., the area where the spatial distribution density of the intersection points of the reverse vectors exceeds a preset distribution density threshold. This convergence region is identified as the danger source area because the animal may have escaped from this area and exhibited abnormal behavior. For example, the activity trajectories of goats and squirrels both show that they moved rapidly from a certain area on the mountainside to the foot of the mountain. After reverse tracing, it was found that their movement vectors converged to a small area on the mountainside, which is the danger source area. Afterwards, the system spatially compares the location information of the identified landslide risk area with the danger source area to find the overlapping area. This overlapping area is both a geologically weak zone and a potential source of risk that could trigger abnormal animal behavior, therefore the risk of landslide is extremely high. Finally, based on the location information of the overlapping area, the system controls the pre-deployed geological sensors within that area to initiate data acquisition. These sensors may include displacement sensors, stress sensors, crack monitors, and soil moisture sensors, which collect geological parameters of the overlapping area in real time, such as minute displacements on the mountain surface, stress changes within the rock, crack propagation, and soil moisture. During execution, the following details need attention: When obtaining behavioral correlation scores, ensure the integrity and accuracy of the data to avoid misjudging the highest-scoring animal due to individual abnormal data; perform time synchronization and spatial calibration of activity trajectories, considering time deviations and coordinate system issues between different monitoring devices, which can be achieved through GPS clock synchronization and coordinate transformation algorithms; when identifying hazard source areas, the preset distribution density threshold should be reasonably set based on historical data and animal behavioral characteristics; for example, the threshold may differ between social and solitary animals; when controlling the geological sensors to collect data, confirm that the sensors are functioning normally and that communication links are unobstructed to ensure the real-time nature and reliability of the collected data.In addition, if there are multiple animal pairs with similar behavioral correlation scores, the system can comprehensively consider factors such as the animals' infrasound perception sensitivity and habitat stability, and select animal pairs with greater early warning value for analysis to ensure more accurate location of danger source areas.

[0056] S107. If the geological data shows an anomaly, it is determined that a landslide is about to occur, and an early warning information on abnormal animal behavior is generated based on the distinguishing feature.

[0057] "Geological data" refers to various data related to the geological conditions of a mountain, collected by geological sensors pre-installed in the mountain area. This data covers multiple aspects such as mountain displacement, stress, strain, groundwater level, and soil moisture, reflecting the current geological state of the mountain. "Abnormal animal behavior warning information" refers to alerts generated based on abnormal animal behavior, reminding relevant personnel of a potential landslide. This information typically includes the abnormal animal behavior, the area where a landslide might occur, and the estimated time of the landslide. For example, if, in a certain mountain area, geological sensors collect data on mountain displacement that increases significantly within a short period, clearly exceeding the normal fluctuation range, this constitutes an abnormal geological data event. If, at this time, animals exhibit abnormal behaviors such as restlessness or large-scale migration, a message reminding nearby personnel to evacuate based on these behaviors would be considered an abnormal animal behavior warning.

[0058] Specifically, this step is executed after analyzing the behavioral characteristics of the target resident animal, identifying abnormal behavior (i.e., a matching score greater than a set matching score threshold), and then controlling the geological sensors to collect geological data. The scenario primarily targets mountain areas prone to landslides, aiming to detect dangers early and ensure the safety of people's lives and property and the stability of the ecological environment. In practice, it is first necessary to obtain a set of geological data features from historical landslides. This feature set is obtained by analyzing, summarizing, and generalizing geological data collected from numerous past landslide cases, and it includes typical characteristic patterns exhibited by geological data before different types of landslides occur.

[0059] Then, the currently collected geological data is matched with a geological data feature set. During this matching process, a series of specialized data analysis algorithms and models are employed. For example, for mountain displacement data, the similarity between the current data and the displacement data in the feature set is calculated, evaluating the degree of similarity by comparing multiple dimensions such as the magnitude, speed, and trend of displacement changes. For stress data, the distribution, variation patterns, and differences from the stress data in the feature set are analyzed. Through comprehensive analysis of various geological data, a mountain collapse matching score is obtained, reflecting the degree of matching between the current geological data and the historical mountain collapse geological data feature set. If this mountain collapse matching score is greater than a set mountaintop collapse matching score threshold, it can be determined that the current geological data is abnormal, meaning the mountain is in an unstable state and a mountain collapse is imminent. Once an impending mountain collapse is determined, the system generates abnormal animal behavior warning information based on the previously acquired distinguishing characteristics of the target resident animals. When generating early warning information, the system details unusual animal behaviors, such as animals that normally frequent fixed areas suddenly migrating in different directions, or certain animals exhibiting restlessness and frequent running. It also clearly identifies potential landslide areas, determined by combining data on the concentration of abnormal animal behavior with the geological structure of the mountain. Furthermore, it provides a preliminary estimate of the expected landslide time. While accurately predicting the landslide time is challenging, a general timeframe can be provided based on the trends and severity of geological data changes. Finally, this crucial information is integrated and promptly and accurately sent to people in potentially affected areas through various channels, including SMS, broadcasts, and alarm systems, enabling them to quickly take countermeasures, evacuate to safety, and minimize losses from the landslide.

[0060] In this embodiment, a series of technical means are employed, including acquiring animal data of the target mountain area to screen for target resident animals, determining monitoring points based on their activity range and monitoring them, identifying distinguishing features by comparing real-time and historical behavioral characteristics of the animals, matching the distinguishing features to obtain a matching score, controlling geological sensors to collect geological data when the matching score exceeds a threshold, and judging the risk of landslides based on the geological data. Therefore, this approach can fully utilize the sensitivity of animals to geological changes and the accurate monitoring of mountain conditions by geological sensors, effectively solving the problems of traditional landslide prediction methods that are difficult to detect mountain anomalies in the early stages and have insufficient prediction accuracy and timeliness. This enables early warning of landslides, ensuring the safety of life and property of people around the mountain and the stability of the ecological environment, and providing strong support for disaster prevention and mitigation work.

[0061] Please refer to the following in conjunction with the above content. Figure 2 The geologically weak zones in the aforementioned areas of abnormal animal behavior are described in detail. Figure 2This is a schematic diagram of an application scenario of the landslide prediction method in the embodiments of this application.

[0062] Assuming the target resident animals in the target mountain area are rabbits and birds, geological structure data for the area has been pre-acquired, including information on potential fault zones, weathered rock layers, and other geologically weak areas. Real-time monitoring of rabbit and bird behavior is conducted at pre-set monitoring points. If a geological disaster is imminent, rabbits and birds will exhibit obvious abnormal behavior in the period leading up to it: rabbits will flee in different directions, but all will move away from potential disaster areas, while flocks of birds will fly away in groups in the direction of the arrows. These abnormal animal behaviors are collected and their spatial location data is used to construct a spatial sequence. Density clustering analysis of this spatial sequence reveals that the abnormal activity of rabbits and birds is concentrated in a certain area, forming a cluster of abnormal animal behaviors. Based on the spatial boundary of this cluster, the area encompassing these abnormal animal behaviors is determined. Subsequently, this area of ​​abnormal animal behavior is spatially overlaid with the mountain's geological structure data. The areas where the area of ​​abnormal animal behavior overlaps with the weak parts such as fault zones and weathered rock layers in the geological structure data are identified as the geologically weak zones of the mountain.

[0063] Below Figure 2 Based on this, we can further pinpoint the source of the hazard at the address, facilitating more precise data collection in the future. Please refer to the following. Figure 3 The dangerous source areas of the geologically weak zones are identified. Figure 3 This is a schematic diagram of another application scenario of the landslide prediction method in the embodiments of this application.

[0064] exist Figure 3 (a) illustrates the spatial distribution and trajectory distribution of abnormal animal behavior within the target mountain area. The system, through recorded video, can identify multiple abnormal activity trajectory points (such as L3, L4, A, C, etc.) of resident animals within a specific time period: rabbits escape from point A to a safe area towards point B via path L1, while another group of rabbits escapes from point E to a safe area towards point F via path L2; similarly, birds escape from a safe area towards point D via path L3, while another group of birds flies away from a safe area towards point H via path L4. These abnormal animal activities all occur during this time period. Figure 2 The resulting weak geological zones, however, if these zones are too broad, it will hinder more precise subsequent geological testing. Therefore, it is possible to... Figure 3 The method shown in (b) is used to determine more precise hazard source areas in geologically weak zones.

[0065] exist Figure 3 In Figure 3(b), the path lines L1, L2, L3, and L4 are normalized to the same coordinate system. This coordinate system is a two-dimensional plane with the x-axis and y-axis as horizontal spatial coordinates, and implicitly includes the z-axis representing vertical height, together forming a three-dimensional spatial coordinate system. Reverse motion vectors are extended along each path line, pointing from different directions towards the hazard source area. Then, the convergence region of each reverse motion vector in three-dimensional space is calculated. The convergence region is defined as the area where the spatial distribution density of the intersection points of the reverse vectors exceeds a preset distribution density threshold. In Figure 3(b), intersection points 1, 2, and 3 are the points formed by the intersection of different reverse motion vector extensions. By statistically analyzing a large number of intersection points and calculating their spatial distribution density, if the distance between two adjacent intersection points is less than a set distance threshold, the preset range area of ​​the intersection point connection can be determined as the convergence region; or if the intersection point density of a certain area exceeds the preset density threshold, that area can be determined as the convergence region. Finally, the convergence region is determined as the hazard source area. In Figure 3(b), the area marked by the gray ellipse is the danger source area identified through the above analysis. This area has a large number of intersections of opposite movement vectors, indicating that the abnormal animal behavior is likely caused by geological activity in this area. Comparison of this area with known geologically weak zones reveals a high degree of overlap, further validating the rationale for identifying this area as a danger source region.

[0066] Based on the above, the following is a more detailed description of the process provided in this implementation. Please refer to [link / reference]. Figure 4 This is another flowchart illustrating the landslide prediction method in this application embodiment.

[0067] S401. Obtain the time series and spatial series of each distinguishing feature;

[0068] Among them, time series refers to the sequence formed by arranging distinguishing features in chronological order of occurrence, used to record the distribution of distinguishing features in the time dimension. For example, if an animal makes abnormal noises at 8 am and moves frequently at 9 am, these events arranged in chronological order constitute a time series. Spatial series refers to the distribution sequence of distinguishing features in spatial location, used to reflect the specific location and movement path of abnormal animal behavior. For example, the location change sequence formed by an animal moving from habitat A to area B in the northwest direction.

[0069] Specifically, this step is executed after comparing the real-time and historical behavioral characteristics of each target resident animal and identifying distinguishing features. The scenario involves the collapse prediction system analyzing and processing animal behavior data. The system needs to analyze each distinguishing feature across time and space. First, the system extracts the occurrence time points corresponding to each distinguishing feature from the real-time video data acquired from monitoring equipment, such as the moment an animal suddenly runs or the time of abnormal gathering. These time points are then arranged into a time series, clearly showing the temporal sequence and distribution patterns of different distinguishing features. Simultaneously, for each distinguishing feature, the system combines the spatial coordinates of the monitoring points to determine the specific location where the feature occurred. For example, if a deer abnormally wanders at coordinates (X1, Y1) and then moves to (X2, Y2), these location information connected chronologically form a spatial sequence. During this process, the system needs to ensure the accuracy of the time data, which may involve calibrating the monitoring equipment's time to avoid sequence corruption due to time errors. For spatial location acquisition, GPS positioning technology or other spatial positioning methods may be required to ensure the accuracy of the spatial coordinates of each distinguishing feature. Furthermore, when multiple distinguishing features exist, the system needs to construct temporal and spatial sequences for each feature separately. For example, for distinguishing features 1 and 2 of the target resident animal A, corresponding time and spatial sequences are generated separately for subsequent analysis of different features. Moreover, when processing time sequences, the system also considers the precision of the time—whether it's accurate to the second, minute, or hour—depending on the frequency of monitoring data collection and the needs of subsequent analysis. For spatial sequences, spatial coordinate transformation may be involved, unifying the coordinates acquired by different monitoring devices into the same coordinate system to ensure the consistency and comparability of the spatial sequences.

[0070] Perform the following operation on any two animals from the resident animals of each target:

[0071] S402. Based on the time series, calculate the time interval between the first target resident animal and the second target resident animal, and construct the first activity trajectory of the first target resident animal and the second activity trajectory of the second target resident animal based on the spatial sequence.

[0072] Here, "first target resident animal" and "second target resident animal" refer to two different animals randomly selected from multiple target resident animals for pairwise behavioral correlation analysis. For example, goats and squirrels are selected as the analysis objects. "Time interval" refers to the difference between the occurrence time of the distinguishing features of the first target resident animal and the occurrence time of the distinguishing features of the second target resident animal. It is used to measure the temporal relationship between the two behaviors. For example, if a goat exhibits abnormal movement at 8:00 and a squirrel exhibits abnormal calling at 8:05, the time interval is 5 minutes. "First activity trajectory" is a movement path constructed based on the spatial sequence of the distinguishing features of the first target resident animal. It is a visual representation of its abnormal behavior in space. For example, the path connecting the goat's movement from point A to point B and then to point C on a hillside. Similarly, "second activity trajectory" is a movement path constructed based on the spatial sequence of the distinguishing features of the second target resident animal.

[0073] Specifically, this step is performed after acquiring the time and spatial sequences of each distinguishing feature, in the context of the system performing pairwise behavioral correlation analysis on target resident animals. The system needs to operate on each pair of target resident animals. First, for the selected first and second target resident animals, the system extracts the occurrence times of the distinguishing features from their respective time sequences. For example, suppose the first target resident animal has three distinguishing features in its time sequence, occurring at times t1, t2, and t3, and the second target resident animal has two distinguishing features, occurring at times t4 and t5. The system calculates the time intervals between each time point of the first target resident animal and each time point of the second target resident animal, i.e., |t1-t4|, |t1-t5|, |t2-t4|, |t2-t5|, |t3-t4|, and |t3-t5|. When calculating the time intervals, attention must be paid to the consistency of time; it may be necessary to convert time data from different sources into a unified standard time format to avoid calculation errors caused by time zone, clock errors, and other issues. Simultaneously, the system records these time intervals for later comparison with preset time window thresholds. Next, the activity trajectory is constructed based on the spatial sequence. For the first target resident animal, the system connects the coordinate points in its spatial sequence sequentially according to time to form the first activity trajectory. For example, if the coordinate points in the spatial sequence are (x1, y1, z1), (x2, y2, z2), and (x3, y3, z3), then the first activity trajectory is the curve or polyline formed by connecting these three points sequentially. When constructing the trajectory, it may be necessary to smooth the spatial coordinates to eliminate trajectory jitter caused by positioning errors, making the trajectory more reflective of the animal's actual movement path. The same method is used to construct the second activity trajectory for the second target resident animal. In addition, when constructing the activity trajectory, the time dimension also needs to be considered to ensure that each point on the trajectory corresponds to an accurate timestamp, so as to analyze the time change pattern of the trajectory later. When the activity trajectory of the target resident animal involves three-dimensional space, the system needs to have the ability to perform three-dimensional modeling, incorporating the height information (such as the z-axis) in the spatial coordinates into the trajectory construction to more comprehensively reflect the animal's activity range. Moreover, different species of animals may have different activity patterns. For example, the activity patterns of birds may vary greatly in the vertical direction, while the patterns of terrestrial animals are mainly on the horizontal plane. The system needs to perform corresponding trajectory processing and analysis based on the characteristics of different animal species.

[0074] S403. Compare the time interval with the preset time window threshold to determine the time synchronization score;

[0075] Here, the time interval represents the difference in the time of occurrence of the distinguishing features between the first and second target resident animals. The preset time window threshold is a pre-defined time range standard used to judge whether the behaviors of the two animals are synchronous in time. For example, it is set to 10 minutes. When the time interval is within this range, the behaviors of the two animals are considered to have a certain degree of synchronicity in time. The time synchronicity score is a score determined based on the comparison between the time interval and the preset time window threshold. It is used to quantify the degree of correlation between the behaviors of the two animals in the time dimension. The higher the score, the stronger the time synchronicity.

[0076] Specifically, this step is executed after calculating the time interval between the first and second target resident animals, during the system's analysis of the temporal correlation of pairwise animal behaviors. The system needs to evaluate each calculated time interval. First, the system retrieves a pre-set time window threshold, which may be derived from historical data statistical analysis. For example, by studying the time intervals of animal behaviors under normal conditions, it may be determined that when the time interval is less than or equal to 10 minutes, there may be a certain correlation in animal behaviors. Then, each time interval is compared with this threshold. For example, if the time interval is 5 minutes, less than the 10-minute threshold, it indicates that the distinguishing features of the two animals occur at relatively close times, showing temporal synchronicity; if the time interval is 15 minutes, greater than the threshold, it indicates poor temporal synchronicity. The system may use different scoring rules when determining the temporal synchronicity score. A common approach is linear scoring, where the score increases as the time interval decreases when it is less than or equal to the threshold. For example, if the threshold is 10 minutes, the score is 100 points when the time interval is 0 minutes, 60 points when the time interval is 10 minutes, and 0 points when the time interval exceeds 10 minutes. Another approach is segmented scoring, which divides the time interval into different ranges, each range corresponding to a different score. For example, 0-5 minutes gets 80-100 points, 5-10 minutes gets 60-80 points, and more than 10 minutes gets 0 points.

[0077] The system may also consider the direction of the time interval, i.e., whether the behavior of the first target resident animal precedes or follows that of the second target resident animal. In some cases, different time sequences may reflect different behavioral association patterns, and therefore may be given different weights in the scoring. In addition, when there are multiple time intervals, the system needs to determine how to integrate the scores of these intervals to obtain the final temporal synchronicity score.

[0078] S404. Calculate the overlap area and spatial position deviation of the first and second activity trajectories to determine the spatial consistency score;

[0079] The overlapping area refers to the size of the area jointly covered by the first and second activity trajectories in space, used to measure the degree of overlap between the two trajectories in spatial location. For example, if both trajectories pass through a valley area, that area is the overlapping part. Spatial position deviation refers to the average distance or degree of deviation between the two trajectories, reflecting their degree of separation in space. The distance between two points is a measure of spatial position deviation. The spatial consistency score is a score calculated based on the overlapping area and spatial position deviation, used to evaluate the similarity and correlation between the two trajectories in the spatial dimension. The higher the score, the stronger the spatial consistency.

[0080] Specifically, this step is performed after the first and second activity trajectories have been constructed, during the system's analysis of the spatial behavioral correlations of the target resident animals. The system needs to perform detailed calculations and evaluations of the activity trajectories of each pair of target resident animals. First, the overlap area is calculated. The system projects the first and second activity trajectories onto the same plane (usually a horizontal ground) and then determines their coverage area on that plane. This may involve discretizing the trajectories into a series of points and then using a polygon region recognition algorithm to determine their respective coverage areas. For example, for a continuous activity trajectory, it can be divided into multiple line segments, each connecting two adjacent coordinate points; the area enclosed by these line segments is the coverage area of ​​the trajectory.

[0081] The system then identifies the intersection of the two coverage areas and calculates the area of ​​this intersection as the overlap area. During this calculation, the time dimension of the trajectories may need to be considered; for example, only the overlapping portion of the two trajectories within the same time period may be calculated, as animal activity can vary over time. Next, the spatial positional deviation is calculated. The system selects a series of corresponding points on the two trajectories, typically choosing locations at the same or close time points in chronological order. Then, the distance between each pair of corresponding points is calculated, such as the Euclidean distance. Finally, the average or weighted average of all these distances is taken to obtain the spatial positional deviation. When selecting corresponding points, temporal synchronization is crucial. If the time series of the two trajectories are not perfectly consistent, time interpolation or sampling may be necessary to ensure reasonable correspondence. When determining the spatial consistency score, the system considers both the overlap area and the spatial positional deviation. A common approach is to use the overlap area as a positive indicator and the spatial positional deviation as a negative indicator, calculating the score through a weighted summation.

[0082] Another approach is to normalize the overlap area and spatial position deviation separately, and then take a weighted average of the two. Furthermore, the system may consider the shape similarity of the trajectories, for example, using shape matching algorithms to assess whether the overall shapes of two trajectories are similar, incorporating shape similarity into the spatial consistency score calculation. Different calculation methods may be used for different types of animal activity trajectories. For example, for the three-dimensional trajectories of flying animals, it may be necessary to calculate the overlap volume and spatial position deviation in three-dimensional space; for the trajectories of ground animals, calculations are mainly performed in a two-dimensional plane. Simultaneously, the system also needs to consider the density and complexity of the trajectories; for complex trajectories, simplification may be necessary to improve computational efficiency and accuracy.

[0083] S405. Based on the temporal synchronicity score and spatial consistency score, calculate the behavioral correlation score between the first target resident animal and the second target resident animal;

[0084] Among them, the behavioral correlation score is the final score obtained by comprehensively considering temporal synchronicity and spatial consistency, and is used to comprehensively assess the strength of the correlation between the behaviors of the two animals.

[0085] Specifically, this step is performed after the temporal synchronicity score and spatial consistency score have been determined, during the system's comprehensive assessment of the behavioral correlations between pairs of target resident animals. The system needs to integrate the scores from these two dimensions. First, the system determines the weights of the temporal synchronicity score and the spatial consistency score. The weighting of these two weights depends on the specific application scenario and research objective. For example, in some cases, temporal synchronicity may be more important because the simultaneous occurrence of animal behaviors may better reflect their response to the same external stimulus; in this case, the weight of the temporal synchronicity score can be set higher. In other cases, spatial consistency may be more critical because the activity of animals in close spatial areas may indicate that they are influenced by the same environmental factors; in this case, the weight of the spatial consistency score can be set higher. A common method for determining the weights is based on analysis of expert knowledge or historical data. For example, through research on animal behavior before previous landslide events, it was found that the contribution ratio of temporal synchronicity and spatial consistency to landslide prediction is 6:4; therefore, the weight of the temporal synchronicity score can be set to 0.6, and the weight of the spatial consistency score can be set to 0.4. Then, the system calculates the behavioral correlation score using a weighted summation method.

[0086] S406. Integrate the behavioral correlation scores between all pairs of animals in each target resident animal to obtain the matching score.

[0087] Specifically, this step is performed after calculating the behavioral correlation scores between all pairs of target resident animals, during the system's comprehensive evaluation of the behavior of the entire target resident animal group. The system needs to integrate these scores. First, the system collects the behavioral correlation scores between all pairs of animals. Then, the system uses an appropriate method to integrate these scores. A common method is to calculate an average. The behavioral correlation scores between all pairs of animals are summed and then divided by the total number of combinations to obtain the average behavioral correlation score as the matching score. Another method is to calculate a weighted average. Different weights are assigned to each behavioral correlation score based on the importance or reliability of different animal pairs. For example, animal pairs with higher infrasound perception sensitivity may be given higher weights because their behavioral changes may better reflect potential changes in the mountain. The system may also consider the distribution of the behavioral correlation scores. For example, statistics such as the median, mode, or variance of the scores are calculated to gain a more comprehensive understanding of the score distribution characteristics. If the variance of the scores is small, it indicates that the behavioral correlation between animal pairs is relatively consistent, and the matching score may be higher; conversely, if the variance is large, it indicates that there is a large difference in the behavioral correlation, and the matching score may be lower.

[0088] In this embodiment, the technical solution employs the acquisition of time and spatial sequences of various distinguishing features, and based on this, calculates time intervals between animals, constructs activity trajectories, determines temporal synchronicity scores and spatial consistency scores, and then calculates behavioral correlation scores and integrates them to obtain matching scores. Therefore, it can comprehensively and accurately analyze the correlation and consistency of the behavior of the target resident animal group, effectively solving the problem that it is difficult to accurately judge potential risks through animal behavior in traditional landslide prediction, and thus achieving the technical effect of using abnormal animal behavior to provide early warning of landslides.

[0089] The collapse prediction system in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference]. Figure 5 This is a schematic diagram of the physical device structure of a collapse prediction system in an embodiment of this application.

[0090] It should be noted that, Figure 5 The structure of the collapse prediction system shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0091] like Figure 5As shown, the collapse prediction system includes a Central Processing Unit (CPU) 501, which can perform various appropriate actions and processes based on a program stored in Read-Only Memory (ROM) 502 or a program loaded from storage section 508 into Random Access Memory (RAM) 503, such as performing the methods described in the above embodiments. The RAM 503 also stores various programs and data required for system operation. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An Input / Output (I / O) interface 505 is also connected to the bus 504.

[0092] The following components are connected to I / O interface 505: input section 506 including audio input devices, push-button switches, etc.; output section 507 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 508 including a hard disk, etc.; and communication section 509 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 509 performs communication processing via a network such as the Internet. Drive 510 is also connected to I / O interface 505 as needed. Removable media 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 510 as needed so that computer programs read from them can be installed into storage section 508 as needed.

[0093] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit (CPU) 501, it performs the various functions defined in the present invention.

[0094] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0096] Specifically, the landslide prediction system of this embodiment includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements the landslide prediction method provided in the above embodiment.

[0097] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the landslide prediction system described in the above embodiments; or it may exist independently and not assembled into the landslide prediction system. The storage medium carries one or more computer programs, which, when executed by a processor of the landslide prediction system, cause the landslide prediction system to implement the landslide prediction method provided in the above embodiments.

[0098] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0099] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0100] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for predicting landslides, characterized in that, The method includes: Obtain animal data for the target mountain area, and filter the animal data to identify multiple different target resident animals. The animal data shall include at least the animals' daily activity range. Monitoring points are determined based on the daily activity range of each of the target resident animals, and each of the target resident animals is monitored based on the monitoring points. Acquire historical daily activity video data of each of the target resident animals within a set time period, and extract the corresponding historical behavioral features of each of the target resident animals; acquire real-time daily activity video data of each of the target resident animals, and extract the corresponding real-time behavioral features of each of the target resident animals. The real-time behavioral characteristics of each of the target resident animals are compared with the historical behavioral characteristics to determine the distinguishing features; The distinguishing features of each of the target resident animals are matched to obtain a matching score; If the matching score is greater than the set matching score threshold, then control multiple pre-set geological sensors to collect geological data; If the geological data shows an anomaly, it is determined that a landslide is about to occur, and an early warning information on abnormal animal behavior is generated based on the distinguishing features. The step of matching the distinguishing features of each of the target resident animals to obtain a matching score includes: Obtain the time series and spatial series of each of the distinguishing features; Perform the following operation on any two animals from the target resident animals: Based on the time series, the time interval between the first target resident animal and the second target resident animal is calculated, and based on the spatial series, a first activity trajectory of the first target resident animal and a second activity trajectory of the second target resident animal are constructed. The time interval is compared with a preset time window threshold to determine the time synchronization score; Calculate the overlap area and spatial position deviation of the first and second activity trajectories to determine the spatial consistency score; Based on the time synchronization score and the spatial consistency score, the behavioral correlation scores of the first target resident animal and the second target resident animal are calculated; The matching score is obtained by integrating the behavioral correlation scores between all pairs of animals in each of the target resident animals; Before the step of controlling multiple pre-set geological sensors to collect geological data, the method further includes: Obtain geological structure data of the mountain area; Density clustering analysis was performed on the spatial sequence to construct clusters of abnormal animal behaviors; Based on the spatial boundaries of the aggregation area, the regions exhibiting abnormal animal behavior are identified. Spatial overlay analysis of the areas of abnormal animal behavior and the geological structure data of the mountain is performed to determine the geologically weak zones corresponding to the areas of abnormal animal behavior. The location information of the collapse risk area is determined based on the geologically weak zone; Following the step of determining the location information of the landslide risk area based on the geologically weak zone, the method further includes: Obtain the two target resident animals with the highest scores among the behavioral correlation scores described above; Obtain the activity trajectories of the two target resident animals, and determine the danger source area based on the activity trajectories; Determine the overlapping area between the location information and the hazard source area; The geological sensors located in the overlapping area are controlled to collect geological data.

2. The method according to claim 1, characterized in that, The process involves acquiring animal data for the target mountain area and filtering the data to identify several different target resident animals, including: Obtain the infrasound sensing sensitivity of each animal and the activity range curve of each animal from the animal data; Calculate the perception weight of the infrasound sensing sensitivity and the habitat stability weight of the activity range curve; Combining the perception weight and the stability weight, the top N animals are selected as target resident animals.

3. The method according to claim 2, characterized in that, The step of determining the danger source area based on the activity trajectory includes: The activity trajectory data is synchronized in time and calibrated in space to generate a three-dimensional motion path with timestamps; Trajectory backtracking is performed in the opposite direction of the motion path to construct a reverse motion vector set; Calculate the convergence region of each reverse motion vector in three-dimensional space. The convergence region is the region where the spatial distribution density of the intersection points of the reverse vectors exceeds a preset distribution density threshold. The convergence region is identified as the hazard source region.

4. The method according to claim 1, characterized in that, The monitoring of each of the target resident animals based on the monitoring points includes: Multiple infrared sensors pre-positioned at the monitoring points are used to detect in real time whether there are animal heat source signals in the surrounding preset area environment. When the animal heat source signal is detected, the sound sensor is controlled to collect target sound data; If the target sound data matches the sound frequency characteristics of the target resident animal, then the monitoring equipment corresponding to the monitoring point is triggered to collect video data of the target resident animal. Within a set startup time after the monitoring equipment is started, the collected video data is identified to determine whether the animal corresponding to the animal heat source signal is the target resident animal; If so, continue collecting video data; If not, then control the monitoring equipment to stop working; If the target resident animal leaves the set monitoring range, or if no heat source signal from the animal is detected within a preset waiting time, the monitoring equipment will be controlled to stop working.

5. A collapse prediction system, characterized in that, The collapse prediction system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the collapse prediction system to perform the method as described in any one of claims 1-4.

6. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the collapse prediction system, the collapse prediction system performs the method as described in any one of claims 1-4.

7. A computer program product, characterized in that, When the computer program product is run on the collapse prediction system, the collapse prediction system performs the method as described in any one of claims 1-4.

Images