A method and system for intelligent monitoring of rodents

Abstract

The application relates to the technical field of vector monitoring, and discloses a rodent intelligent monitoring method and system, computer equipment and a storage medium. The method comprises the following steps: acquiring a plurality of trigger events, each trigger event comprising device identification information and a trigger timestamp; querying a device position mapping table to generate a position marking event; constructing a space-time correlation graph based on the plurality of position marking events; inputting the space-time correlation graph into an individual identification model, the model solving the optimal path division of the space-time correlation graph based on a maximum possible moving speed constraint, and outputting a position marking event sequence belonging to the same target rodent; sequentially connecting adjacent position marking events in the sequence in the order of the trigger timestamps, and fitting to generate activity trajectory data; and generating early warning information when the activity trajectory data meets a preset early warning condition. The application can identify individual activity sequences from discrete trigger events and fit to generate continuous trajectories, thereby realizing fine monitoring of rodent individual activities.

Description

Technical Field

[0001] This application relates to the field of vector-borne disease monitoring technology, specifically to an intelligent monitoring method, system, computer equipment, and storage medium for rodents. Background Technology

[0002] With the development of IoT technology, intelligent monitoring methods have begun to be applied in the field of vector-borne disease control, leading to the emergence of technologies for rodent monitoring using sensors. Current monitoring methods primarily involve deploying traps within the monitoring area. When a rodent triggers the trap, the trigger event is recorded, thus obtaining information about the animal's presence. However, in scenarios where multiple monitoring and trapping devices operate simultaneously, traditional monitoring methods can only provide simple information such as "a mouse has appeared in a certain location," failing to distinguish trigger events generated by different individuals or to correlate multiple triggering behaviors of the same target animal to form a continuous movement path. When multiple rodents are active within the monitoring area, the discrete events triggered by different individuals intertwine, making it impossible to accurately track the activity patterns of individual animals. This lack of individual identification capability means that monitoring data can only reflect a rough picture of activity at the group level, making it difficult to achieve refined analysis of rodent activity trajectories. Summary of the Invention

[0003] Therefore, it is necessary to provide an intelligent monitoring method, system, computer equipment, and storage medium that can realize individual rodent identification and activity trajectory tracking to address the above-mentioned technical problems.

[0004] Firstly, a method for intelligent monitoring of rodents is provided, the method comprising: Multiple trigger events are acquired. These multiple trigger events are generated by multiple monitoring and capture devices deployed in a preset monitoring area in response to the entry of a target rodent. Each trigger event includes device identification information and a trigger timestamp. Based on the device identification information, a preset device location mapping table is queried, and the triggering event is mapped to the corresponding location on a preset monitoring area map to generate a location marker event containing location information and a trigger timestamp; Based on multiple location marker events, a spatiotemporal association graph is constructed with the location marker events as nodes and the order of their trigger timestamps as edges; The spatiotemporal correlation graph is input into the individual identification model, which solves the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible movement speed constraint, and outputs the location marker event sequence belonging to the same target rodent. According to the order of the trigger timestamps, adjacent location marker events in the location marker event sequence belonging to the same target rodent are connected in sequence to fit and generate the activity trajectory data of the target rodent; When the activity trajectory data meets the preset warning conditions, a warning message is generated.

[0005] In one embodiment, the maximum possible moving speed constraint is determined in the following way: Obtain the velocity probability distribution function obtained by statistical learning based on all trigger events within a preset historical time period. The velocity probability distribution function represents the probability distribution of the movement speed of the target rodent under a given time interval. Based on the speed probability distribution function and a preset confidence level, the maximum possible moving speed corresponding to any given time interval is determined, wherein the speed value corresponding to the maximum possible moving speed satisfies the following condition: the cumulative probability that the moving speed is less than or equal to the speed value is equal to the confidence level.

[0006] In one embodiment, obtaining the velocity probability distribution function obtained through statistical learning of all trigger event pairs within a preset historical time period includes: Extract all trigger event pairs within a preset historical time period from the database. Each trigger event pair includes the time interval and spatial distance corresponding to two consecutive triggers by the same monitoring and capture device or the sequential triggers by different monitoring and capture devices. Calculate the ratio of the spatial distance to the time interval for each of the trigger event pairs, and use it as the instantaneous movement speed corresponding to each of the trigger event pairs; Based on the magnitude of the time interval, the instantaneous movement speed is divided into multiple preset time interval intervals, each of which corresponds to a continuous time range; Kernel density estimation is performed on all instantaneous movement velocities within each time interval to obtain the velocity probability density function within that time interval; By parametrically fitting the velocity probability density function within different time intervals, a continuous velocity probability distribution function is obtained.

[0007] In one embodiment, the confidence level is dynamically adjusted based on captured verification feedback, the dynamic adjustment including: When a capture confirmation signal is received from the monitoring and capture device, the capture confirmation signal is associated with the corresponding triggering event to obtain positive samples that truly belong to the same target rodent and negative samples that truly belong to different target rodents. Based on the positive and negative samples, calculate the recall and precision at the current confidence level; When the recall rate is lower than a preset recall rate threshold, the confidence level is reduced; when the precision rate is lower than a preset precision rate threshold, the confidence level is increased.

[0008] In one embodiment, the step of constructing a spatiotemporal relationship graph based on multiple location marker events, with the location marker events as nodes and the order of their trigger timestamps as edges, includes: Using each location marker event as a node, directed edges are constructed according to the order of the trigger timestamps to obtain an initial directed graph; For any two nodes connected by a directed edge in the initial directed graph, calculate the time interval and spatial distance between the corresponding two position marker events; If the spatial distance is less than or equal to the product of the maximum possible movement speed and the time interval, retain the directed edge between the two nodes; otherwise, delete the directed edge. Each of the retained directed edges is assigned an edge weight, which is a monotonically decreasing function of time interval and spatial distance, with the edge weight between two nodes having a shorter time interval and a closer spatial distance being larger. The set of directed edges that have been weighted is used as the spatiotemporal relation graph.

[0009] In one embodiment, the individual identification model solves for the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible movement speed constraint, and outputs a sequence of location-marked events belonging to the same target rodent, including: The individual identification model is based on the maximum spanning tree algorithm, which searches for the path partition with the maximum edge weight in the spatiotemporal correlation graph. The sequence of location marker events corresponding to the nodes contained in each searched path is taken as the sequence of location marker events belonging to the same target rodent.

[0010] In one embodiment, the method further includes: Based on multiple activity trajectory data, the frequency of rodents passing through the location of each monitoring and capture device is counted. Based on the frequency of passage, the preset monitoring area is divided into a high-frequency activity area, a low-frequency activity area, and an inactive area; Generate control instructions for the high-frequency activity area, including suggestions for increasing the density of the monitoring and capture devices or for placing killing agents.

[0011] Secondly, a rodent intelligent monitoring system is provided, comprising: Multiple monitoring and capture devices are deployed within a preset monitoring area to generate trigger events in response to the entry of a target rodent. The trigger events include device identification information and a trigger timestamp. The event acquisition module is communicatively connected to the plurality of monitoring and capturing devices and is used to acquire the plurality of triggering events; The location mapping module, connected to the event acquisition module, is used to query a preset device location mapping table based on the device identification information, map the triggering event to the corresponding location on a preset monitoring area map, and generate a location marker event containing location information and a trigger timestamp. The graph construction module, connected to the location mapping module, is used to construct a spatiotemporal association graph based on multiple location marker events, with the location marker events as nodes and the order of their trigger timestamps as edges; An individual identification module, connected to the graph construction module, is used to input the spatiotemporal association graph into the individual identification model. The individual identification model solves the optimal path partitioning of the spatiotemporal association graph based on the maximum possible movement speed constraint and outputs a sequence of location marker events belonging to the same target rodent. The trajectory fitting module, connected to the individual identification module, is used to sequentially connect adjacent position marker events in the position marker event sequence belonging to the same target rodent according to the order of the trigger timestamps, and fit and generate the activity trajectory data of the target rodent. The early warning module, connected to the trajectory fitting module, is used to generate early warning information when the activity trajectory data meets preset early warning conditions.

[0012] Thirdly, a computer device is provided, including a memory and a processor, wherein the memory is communicatively connected to the processor, and the memory stores a computer program that can run on the processor, wherein when the processor executes the computer program, it implements the intelligent rodent monitoring method described above.

[0013] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the above-described intelligent rodent monitoring method.

[0014] The aforementioned intelligent rodent monitoring method, system, computer equipment, and storage medium acquire trigger events generated by multiple monitoring and capture devices. Based on device identification information, these trigger events are mapped to corresponding locations on a monitoring area map to generate location marker events. A spatiotemporal correlation graph is then constructed. This graph is input into an individual identification model, which solves for the optimal path division based on the maximum possible movement speed constraint, outputting a sequence of location marker events belonging to the same target rodent. Adjacent location marker events are then sequentially connected and fitted according to their trigger timestamps to generate activity trajectory data. Finally, an early warning message is generated when the activity trajectory data meets preset warning conditions. This allows for accurate identification of trigger event sequences belonging to the same target rodent from discrete trigger events generated by multiple monitoring and capture devices without the need for additional hardware. By fitting and generating continuous activity trajectory data, it achieves refined tracking of individual animals, solving the technical problem of difficulty in associating and attributing discrete events in group activity scenarios, and improving the accuracy and early warning capability of rodent monitoring. Attached Figure Description

[0015] Figure 1 This is a diagram illustrating the application environment of an intelligent monitoring method for rodents in one embodiment. Figure 2 This is a flowchart illustrating an intelligent monitoring method for rodents in one embodiment; Figure 3 This is a flowchart illustrating the process of determining the maximum possible moving speed constraint in one embodiment; Figure 4 This is a flowchart illustrating the process of obtaining the velocity probability distribution function in one embodiment; Figure 5 This is a flowchart illustrating the dynamic adjustment of confidence levels in one embodiment; Figure 6 This is a flowchart illustrating the process of constructing a spatiotemporal correlation graph in one embodiment; Figure 7 This is a flowchart illustrating the process of dividing the optimal path in one embodiment; Figure 8 This is a flowchart illustrating the process of dividing an activity area and generating prevention and control instructions in one embodiment; Figure 9 This is a structural block diagram of a rodent intelligent monitoring system in one embodiment; Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0016] To facilitate understanding of the technical solutions provided in the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0017] In the field of vector-borne disease control, rodents (such as rats) are the main vectors for many infectious diseases, including plague and hemorrhagic fever with renal syndrome. Monitoring their activity patterns is a fundamental task in disease prevention and control. Traditional rodent monitoring methods mainly employ overnight traps and cage traps: overnight traps are set up, and rats are captured and counted the following day; cage traps are used to capture live rats for pathogen detection and investigation of external parasites. Both methods are simple to operate and inexpensive, but they rely on manual placement and retrieval, resulting in high labor intensity, low work efficiency, and difficulty in achieving continuous monitoring.

[0018] With the development of IoT technology, intelligent monitoring methods are beginning to be applied in the field of vector-borne disease control. Existing technologies include intelligent rodent traps based on sensors and wireless communication. When a rodent enters the device and triggers the sensor, the device generates a trigger event and sends it to a server via a wireless network, recording the animal's appearance time and location information. These intelligent devices achieve automatic data collection and remote transmission, improving efficiency and real-time performance compared to traditional manual methods.

[0019] However, in practical applications, the monitoring area is typically equipped with multiple intelligent rodent traps, forming a distributed monitoring network. When multiple rodents move within the area, the devices triggered by different individuals generate numerous discrete triggering events. For example, a mouse foraging at night may trigger multiple devices sequentially, while the activities of another mouse may intertwine with the triggering events of the previous mouse in time and space. Although existing technologies can record the time and location of each triggering event, they lack effective individual identification methods, making it impossible to distinguish which triggering events were caused by the same mouse, or to link multiple triggering behaviors of the same mouse to form a continuous movement path. The resulting monitoring data is merely an isolated event record, reflecting only group-level information such as "a mouse appeared at a certain location at a certain time," and cannot achieve fine-grained tracking of the activity patterns of individual animals.

[0020] This lack of individual identification capability severely limits the analytical value of monitoring data. For example, when assessing the frequency of rodent activity in a certain area, current technology cannot distinguish between repeated appearances of the same rodent and separate appearances of multiple rodents; when tracking rodent migration routes, current technology cannot connect discrete trigger points into continuous trajectories; and when identifying rodent habitats, current technology cannot perform spatial clustering analysis based on activity patterns. Therefore, accurately identifying trigger event sequences belonging to the same target rodent from discrete trigger events generated by multiple monitoring and capture devices, and fitting them to generate continuous activity trajectory data, has become a pressing technical problem to be solved in this field.

[0021] Therefore, this embodiment provides a method, system, computer equipment, and storage medium for intelligent monitoring of rodents.

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. In the implementation of the embodiments of this application, all data collection and use have been authorized and comply with privacy regulations.

[0023] The intelligent monitoring method for rodents provided in this application can be applied to, for example... Figure 1 In the application environment shown, multiple monitoring and capture devices 102 are deployed in a preset monitoring area (e.g., residential areas, warehouses, sewers, etc.) and send trigger events to a server 104 via a network. After receiving multiple trigger events, the server 104 executes the monitoring method provided in this embodiment, including: querying a preset device location mapping table based on device identification information, mapping the trigger events to corresponding locations on a preset monitoring area map to generate location marker events; constructing a spatiotemporal correlation graph with location marker events as nodes and the order of trigger timestamps as edges based on multiple location marker events; inputting the spatiotemporal correlation graph into an individual identification model, which solves for the optimal path division of the spatiotemporal correlation graph based on the maximum possible movement speed constraint, and outputs a sequence of location marker events belonging to the same target rodent; connecting adjacent location marker events in the sequence of location marker events belonging to the same target rodent according to the order of trigger timestamps to fit and generate activity trajectory data of the target rodent; and generating a warning message when the activity trajectory data meets preset warning conditions. Server 104 sends the generated warning information to terminal 106 via the network, and terminal 106 displays the warning information to the user. The monitoring and capture device 102 can be a device with event triggering capabilities, such as a smart mousetrap, infrared sensor, or pressure sensor; terminal 106 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices; server 104 can be a standalone server or a server cluster consisting of multiple servers.

[0024] Firstly, in the monitoring of disease vectors, multiple monitoring and capture devices are deployed within the monitoring area. When multiple disease vectors (such as mice) move around, each device generates a large number of discrete trigger events. Traditional technologies can only record these isolated events and cannot correlate device events triggered multiple times by the same disease vector, making it impossible to track the activity trajectory of a single disease vector and failing to meet the needs of refined monitoring. To address this problem, in one embodiment, an intelligent monitoring method for rodents is provided, which can be applied to... Figure 1 The monitoring and capture device 102 and terminal 106 are described below, with rodent monitoring as an example. Figure 2 As shown, the method includes the following steps: Multiple trigger events are acquired. These trigger events are generated by multiple monitoring and capture devices deployed in a preset monitoring area in response to the entry of a target rodent. Each trigger event includes device identification information and a trigger timestamp.

[0025] In this step, the monitoring and capture device refers to a device with trigger event generation functionality. In this embodiment, it can be a smart rodent trap, such as a rodent cage or trap with built-in pressure sensors, infrared sensors, or reed switches. When a target rodent (such as a mouse) enters the device and triggers the sensor, the device immediately generates a trigger event record. This record contains the device's unique identification information (such as an ID number) and a timestamp of the trigger time. Multiple monitoring and capture devices operate simultaneously, continuously sending trigger events to the server via wireless networks (such as LoRa, NB-IoT, Wi-Fi, etc.). The server receives and stores these events.

[0026] Based on the device identification information, a preset device location mapping table is queried, and the triggered event is mapped to the corresponding location on the preset monitoring area map, generating a location marker event containing location information and a trigger timestamp.

[0027] In this step, the server has a pre-set device location mapping table, which records the correspondence between each device identifier and its coordinate location (such as latitude and longitude, planar coordinates) on the monitoring area map. For each received trigger event, the server queries the mapping table based on the device identifier information to obtain the device's location coordinates, and combines these coordinates with the trigger timestamp to generate a location marker event. A location marker event can be represented as a (location, time) tuple, for example... This step transforms discrete triggering events into analyzable data points with spatial location and temporal information.

[0028] Based on multiple location marker events, a spatiotemporal relationship graph is constructed with location marker events as nodes and the order of trigger timestamps as edges.

[0029] In this step, each location marker event is treated as a node in a graph, with node attributes including location and time. Then, a directed edge is created for every two nodes that occur sequentially in time, based on the order of their trigger timestamps. The direction of the edge is from the earlier node to the later node. For example, if event A occurs... Event B occurred and If so, a directed edge is created from A to B. The resulting initial directed graph contains all possible temporal relationships, but many edges may correspond to events of different individuals, requiring further filtering.

[0030] The spatiotemporal correlation graph is input into the individual identification model. The individual identification model solves the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible movement speed constraint, and outputs the location marker event sequence belonging to the same target rodent.

[0031] In this step, the individual identification model employs a graph partitioning algorithm based on the maximum possible movement speed constraint. This algorithm searches for the optimal path partitioning in the graph, ensuring that the node sequence within each path corresponds to the activity of the same target rodent. The model can be implemented using any existing graph partitioning or path search algorithm, such as a greedy path growth algorithm or a maximum flow minimum cut partitioning method. By solving for the optimal path partitioning, the model outputs a set of node sequences, where nodes in each sequence are considered to be triggered sequentially by the same target rodent.

[0032] Based on the order of the trigger timestamps, adjacent location marker events belonging to the same target rodent are sequentially connected to fit and generate the target rodent's activity trajectory data.

[0033] In this step, the geographical locations corresponding to two adjacent location marker events in the location marker event sequence are connected sequentially with straight lines according to the order of trigger timestamps from smallest to largest, thus obtaining a continuous broken line, which is the activity trajectory data of the target rodent. For example, if a mouse triggers devices A, B, and C in sequence, the corresponding location points... , , Then connect sequentially , The resulting broken line This refers to the activity trajectory data of the mouse.

[0034] When the activity trajectory data meets the preset warning conditions, a warning message is generated.

[0035] In this step, when the generated activity trajectory data meets the preset early warning conditions (such as the trajectory entering a key protection area, the trajectory density exceeding the limit, etc.), the server generates early warning information. This information may include trajectory data, location, time, etc., and is pushed to relevant terminal devices to prompt managers to take corresponding prevention and control measures.

[0036] Based on the above, by identifying sequences belonging to the same target individual from discrete triggering events and fitting continuous trajectories, refined monitoring of rodent individual activities is achieved, improving the accuracy and early warning capabilities of vector-borne disease monitoring.

[0037] When using the maximum possible moving speed constraint for individual identification, determining the specific value of this constraint is a crucial issue. Using a fixed empirical value makes it difficult to adapt to the differences in the movement abilities of rodents of different species, seasons, or environments, leading to a decrease in individual identification accuracy. To address this problem, one embodiment provides a method for determining the maximum possible moving speed constraint based on statistical learning. Specifically, as shown... Figure 3 As shown, the maximum possible moving speed constraint is determined in the following way: Obtain the velocity probability distribution function obtained by statistical learning based on all triggered event pairs within a preset historical time period. The velocity probability distribution function represents the probability distribution of the movement speed of the target rodent under a given time interval. Based on the velocity probability distribution function and a preset confidence level, the maximum possible moving speed corresponding to any given time interval is determined, wherein the velocity value corresponding to the maximum possible moving speed satisfies the following condition: the cumulative probability that the moving speed is less than or equal to the velocity value is equal to the confidence level.

[0038] Specifically, first, historical trigger event pairs need to be acquired. The server retrieves all recorded trigger events from the database within a preset historical time period (e.g., the past month) and sorts these events by their occurrence time. Any two trigger events constitute a trigger event pair if they are triggered twice consecutively by the same monitoring device, or triggered sequentially by different monitoring devices. Each trigger event pair includes the time interval between the two events. (Unit: seconds) and spatial distance (Unit: meters) Spatial distance can be obtained by calculating the Euclidean distance from the position coordinates of the two devices. In this way, a series of... The sample.

[0039] For these samples, statistical learning methods can be used to obtain the velocity probability distribution function. Specific methods include, but are not limited to: fitting probability densities after grouping samples by time intervals, or directly estimating the conditional distribution using a parametric regression model.

[0040] Then, a preset confidence level is introduced. , It is a value between 0 and 1, for example This indicates the user's desired maximum possible movement speed. It can cover 95% of the possible speeds. For a given... According to the velocity probability distribution function Solve for the condition that satisfies of ,Right now This is the quantile at that confidence level. Thus, It became a kind of... Related functions.

[0041] In practical applications, when it is necessary to determine whether two triggering events belong to the same target, the difference between them is calculated. and Then according to From function Obtain the corresponding maximum possible movement speed and then compare. and The size relationship. If If the speed is feasible, then it is considered feasible; otherwise, it is considered infeasible.

[0042] Based on the above, the maximum possible movement speed is no longer a fixed value, but is obtained through statistical learning from historical data, and then... Dynamic changes can adaptively reflect the movement capabilities of rodents at different time intervals, thereby improving the accuracy and robustness of individual identification. Simultaneously, the confidence level... It can be adjusted according to actual needs; for example, a higher value can be selected for high-precision requirements. (e.g., 0.99), in general monitoring, 0.95 is selected, which offers high flexibility.

[0043] To obtain an accurate velocity probability distribution function, a specific method for constructing this function from historical data is needed. Therefore, in one embodiment, the velocity probability distribution function is obtained through statistical learning based on all trigger event pairs within a preset historical time period, such as... Figure 4 As shown, it includes: Extract all trigger event pairs within a preset historical time period from the database. Each trigger event pair includes the time interval and spatial distance corresponding to two consecutive triggers by the same monitoring and capture device or the sequential triggers by different monitoring and capture devices. Calculate the ratio of spatial distance to time interval for each trigger event pair, and use it as the instantaneous movement speed for each trigger event pair. Based on the magnitude of the time interval, the instantaneous movement speed is divided into multiple preset time interval intervals, each time interval interval corresponding to a continuous time range; Kernel density estimation is performed on all instantaneous movement velocities within each time interval to obtain the velocity probability density function within that time interval; By parametrically fitting the velocity probability density function within different time intervals, a continuous velocity probability distribution function is obtained.

[0044] Specifically, first, the server reads all triggered events from the database within a preset historical time period (e.g., the last 30 days) and sorts them according to their trigger timestamps. Then, for the sorted event sequence, it iterates to find all possible trigger event pairs. There are two rules for generating trigger event pairs: one is two consecutive triggers from the same monitoring and capture device, i.e., two events from the same device that are temporally adjacent; the other is sequential triggers from different monitoring and capture devices, i.e., events from any two different devices can form an event pair as long as they are temporally adjacent and no other events are inserted. This definition covers the possibility of consecutive triggers from the same mouse. Each event pair records the time difference between the two events. (Unit: seconds) and the spatial distance between the two devices (Unit: meters) The spatial distance is calculated based on the coordinates of the device's location.

[0045] Secondly, for each event pair, calculate the instantaneous movement speed. .here This indicates the average speed of the movement, expressed in meters per second.

[0046] Next, according to the time interval The size of the interval determines the division of all event pairs into different time intervals. For example, the intervals could be set as follows: [0, 10) seconds, [10, 30) seconds, [30, 60) seconds, [60, 120) seconds, [120, 300) seconds, [300, 600) seconds, etc., with each interval covering a continuous time range. Each event pair is determined according to its... The value falls within the corresponding interval. Here, we use... Indicates the first A time interval, for example , And so on.

[0047] Then, for each time interval, the instantaneous velocity values ​​of all event pairs within that interval are used as samples, and the velocity probability density function for that interval is fitted using kernel density estimation. Kernel density estimation is a non-parametric estimation method that can automatically fit a smooth probability density curve based on the sample distribution. Specifically, suppose there are... Speed ​​value Then the kernel density estimation function is: in For kernel functions (such as Gaussian kernels). The bandwidth parameter can be determined using empirical rules or cross-validation. This yields the velocity probability density function for that interval. It can also be expressed in conditional form. .

[0048] Finally, for different time intervals density function Perform parameterized fitting to obtain a continuous function form. Since the velocity distribution often changes with... As the velocity increases, it exhibits a pattern of decreasing mean and changing variance. A suitable parametric model can be used for fitting, such as assuming the velocity follows a log-normal or gamma distribution, and using the distribution parameters (such as mean and variance) for each interval as... The function is used for regression fitting to obtain the continuous velocity probability distribution function. The analytical expression. For example, the mean function can be fitted. Sum of standard deviation function ,but It can be expressed as the probability density function of a log-normal distribution: Based on the above, this embodiment provides a specific implementation method for learning the speed probability distribution from historical data. This distribution can reflect the statistical regularity of the movement speed of rodents at different time intervals, providing a reliable data foundation for determining the maximum possible movement speed.

[0049] The confidence level, as a key parameter in individual identification, directly affects the strictness of the judgment. If the confidence level remains constant, it may fail to adapt to changes in the monitoring environment (e.g., seasonal changes altering rodent activity patterns). In one embodiment, a mechanism for dynamically adjusting the confidence level is provided to enhance the system's adaptability. Specifically, such as... Figure 5 As shown, the confidence level is dynamically adjusted based on the captured verification feedback. The dynamic adjustment includes: When a capture confirmation signal is received from the monitoring and capture device, the capture confirmation signal is associated with the corresponding trigger event to obtain positive samples that truly belong to the same target rodent and negative samples that truly belong to different target rodents. Calculate the recall and precision at the current confidence level based on positive and negative samples; When the recall rate is lower than the preset recall rate threshold, the confidence level is reduced; when the precision rate is lower than the preset precision rate threshold, the confidence level is increased.

[0050] Specifically, firstly, some monitoring and capture devices have a capture function (such as mousetraps). When a target rodent is successfully captured, the device emits a special capture confirmation signal, which includes device identification information and a capture timestamp. Upon receiving this signal, the server can confirm that the device did indeed capture a mouse at a specific time. Since a capture event necessarily corresponds to a triggering event (the trigger generated when the mouse enters the device), the capture confirmation signal can be associated with the device's triggering event at a similar time, forming a positive sample. This confirms that the triggering event indeed belongs to a specific mouse, and that no further events occurred after the mouse was captured.

[0051] Furthermore, for capture events, negative samples can be generated by utilizing the temporal relationships between events. For example, if a device is triggered by another device shortly after capture (less than the minimum possible movement time), since the captured mouse cannot trigger other devices, these subsequent triggers must belong to other mice and can thus be used as negative samples, i.e., event pairs that truly belong to different targets.

[0052] The server periodically (e.g., daily) collects these positive and negative samples verified by the capture confirmation signal. Positive samples refer to two events confirmed to belong to the same target (e.g., the capture event itself may not constitute a pair, but it can be constructed by the events before and after the capture), while negative samples refer to two events confirmed to belong to different targets (e.g., event pairs triggered by other devices within a short period after the capture).

[0053] Then, these samples are used to evaluate the decision performance at the current confidence level. For each sample (consisting of two triggering events, with...) and According to the current individual identification model (including those based on...) Definite The model determines the target as follows: if the model identifies the target as the same, it is counted as "predicted positive"; otherwise, it is counted as "predicted negative". Combining this with the true labels (positive / negative samples), the current... recall rate and accuracy Recall rate is defined as: Precision is defined as: Adjust based on the relationship between recall and precision and preset thresholds. Both the preset recall and preset precision thresholds can be set to 0.9. If the current recall rate is lower than the recall threshold, it indicates that the model is too conservative and has missed many samples that should belong to the same target. In this case, the confidence level should be lowered to ensure that... Increase the confidence level to make the decision criteria more lenient and improve recall. If the current precision is below the precision threshold, it indicates that the model is too aggressive, incorrectly classifying many different targets as the same target. In this case, the confidence level should be increased to make it more lenient. The smaller the value, the stricter the judgment criteria, and the higher the accuracy.

[0054] Based on the above, the confidence level can be adaptively optimized by dynamically adjusting based on actual captured feedback, enabling the individual recognition model to maintain good performance in different environments and improving the robustness and practicality of the system.

[0055] When constructing a spatiotemporal association graph, the initial directed graph contains all possible temporal associations, but many edges may correspond to invalid associations between different individuals. In one embodiment, a method is provided to filter edges and assign weights using a maximum possible movement speed constraint, thereby constructing a spatiotemporal association graph that reflects the likelihood of individual associations. Specifically, such as... Figure 6 As shown, based on multiple location marker events, a spatiotemporal relationship graph is constructed, with location marker events as nodes and the order of trigger timestamps as edges, including: Using each location-marked event as a node, construct directed edges according to the order of their trigger timestamps to obtain an initial directed graph; For any two nodes connected by a directed edge in the initial directed graph, calculate the time interval and spatial distance between the corresponding two position marker events; If the spatial distance is less than or equal to the product of the maximum possible movement speed and the time interval, retain the directed edge between the two nodes; otherwise, delete the directed edge. Each retained directed edge is assigned a weight, which is a monotonically decreasing function of time interval and spatial distance. The weight of the edge between two nodes with shorter time interval and closer spatial distance is larger. The set of directed edges with assigned weights is used as a spatiotemporal relation graph.

[0056] Specifically, firstly, based on all the generated location marker events, they are treated as nodes, each with location coordinates. and timestamp Then sort the nodes according to their timestamps from smallest to largest. For any two nodes... and ,like Then in and Establish a line between them point to The directed edges are thus included, encompassing all possible temporal relationships to form an initial directed graph. This initial directed graph may be very dense, but most of its edges may be unreasonable due to excessively long time intervals or large spatial distances.

[0057] Secondly, for each edge in the initial directed graph, calculate the time interval between the corresponding two position labeling events. and spatial distance Then, based on the currently determined maximum possible movement speed function... Determine whether it satisfies If the condition is met, it means that the movement speed is kinematically possible, and the edge is retained; otherwise, the movement is considered impossible, and the edge is deleted. This step uses the maximum possible movement speed constraint to filter out the vast majority of unreasonable associations.

[0058] Then, for each of the remaining edges, assign an edge weight. The edge weights are designed to be related to... and The monotonically decreasing function, i.e. The smaller The smaller, The larger the value, the higher the probability that the two events belong to the same goal. For example, the edge weight can be defined as: in and The positive weighting coefficient can be adjusted according to the actual scenario; it can also be defined as: in and The coefficient is positive. Here, we only need to ensure that the edge weight is positively correlated with the probability of association.

[0059] Finally, the graph composed of all filtered and weighted directed edges is used as the final spatiotemporal association graph. This graph not only eliminates kinematically impossible associations, but also quantifies the probability of each association through edge weights, providing a foundation for subsequent path partitioning.

[0060] Based on the above, spatiotemporal correlation graphs can effectively reflect the potential individual relationships between events, which reduces the complexity of subsequent calculations and improves the accuracy of individual identification.

[0061] After constructing the spatiotemporal association graph with edge weights, it is necessary to partition the graph to identify the sequence of nodes belonging to the same target individual. One embodiment provides an optimal path partitioning method based on the maximum spanning tree algorithm. Specifically, as shown... Figure 7 As shown, the individual identification model solves for the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible moving speed constraint, and outputs a sequence of location-marked events belonging to the same target rodent, including: The individual identification model is based on the maximum spanning tree algorithm, which searches for the path partition with the maximum edge weight in the spatiotemporal correlation graph; The sequence of location-marked events corresponding to the nodes contained in each search path is taken as the sequence of location-marked events belonging to the same target rodent.

[0062] Specifically, the individual identification model treats the spatiotemporal relationship graph as a weighted directed graph, where nodes represent location-labeled events, and directed edges have weights (indicating the probability that two events belong to the same target). The model's goal is to divide all nodes into several paths (a path is a sequence of nodes where adjacent nodes are connected by directed edges), maximizing the sum of edge weights within each path while ensuring that there are no overlapping nodes between paths (each node belongs to only one path). This is essentially a path partitioning optimization problem in graph theory.

[0063] This embodiment uses the maximum spanning tree algorithm to solve the problem. The specific steps are as follows: First, since the graph is directed, but the direction of individual movement is fixed (time from smallest to largest), the directed graph can be treated as an undirected graph to handle edge weights, because direction does not affect weight. Then, a maximum spanning tree algorithm (such as Kruskal's algorithm or Prim's algorithm) is run on the graph to construct a maximum spanning tree containing all nodes and the edges connecting them, with the total edge weight of the tree being maximized. The characteristics of a spanning tree are connectivity and acyclicity, but path partitioning requires paths (linear chains), so further processing is needed.

[0064] In a maximum spanning tree, each node may have a degree greater than 2, leading to a branching structure. To obtain path partitions, the tree can be pruned or decomposed. A simple approach is to use nodes with a degree greater than 2 as branch points and then break the tree from these points, decomposing it into several paths (chains). A more precise approach is to, based on the maximum spanning tree, traverse all leaf nodes to obtain all leaf node-to-leaf paths, select the non-overlapping paths with the largest sum of edge weights, and continue until all nodes are covered.

[0065] In practice, an iterative approach can be used: starting with the edge with the largest weight in the graph, gradually construct the path, ensuring that no node is repeated within the path. This is similar to solving the maximum weight path cover problem, which can be solved using a greedy algorithm or dynamic programming.

[0066] Regardless of the specific algorithm, the final result is a set of paths, and the sequence of nodes contained in each path is arranged in ascending order of timestamps. The location marker events corresponding to these nodes are considered to be triggered sequentially by the same target rodent.

[0067] Based on the above, the maximum spanning tree algorithm transforms the complex path partitioning problem into a graph optimization problem, which can effectively find the most reasonable individual affiliation among many possible associations, thereby obtaining an accurate sequence of individual events.

[0068] After obtaining individual activity trajectory data, how to utilize this data to guide actual prevention and control work is an important objective of the monitoring system. One embodiment provides a method for dividing activity areas based on trajectory data and generating prevention and control instructions. Specifically, such as... Figure 8 As shown, the method also includes: Based on multiple activity trajectory data, the frequency of rodents passing through the location of each monitoring and capture device was counted. Based on the frequency of passage, the preset monitoring area is divided into a high-frequency activity area, a low-frequency activity area, and an inactive area; Generate control instructions for high-frequency activity areas, including suggestions on increasing the density of monitoring and capture devices or the location of application of insecticides.

[0069] Specifically, firstly, the server accumulates activity trajectory data for multiple target rodents, with each trajectory consisting of a series of location points (corresponding to the locations where triggering events occur). For each location where a monitoring and capture device is located, the total number of times that location is traversed across all trajectories is counted. For example, if a device appears 10 times in the trajectory data, its traversal frequency is 10. These frequencies reflect the activity intensity of rodents at each location.

[0070] Then, the entire preset monitoring area is spatially divided according to frequency. Clustering algorithms (such as K-means) or density-based spatial partitioning methods can be used to divide the continuous area into several sub-regions. Based on the average or highest frequency of devices within each sub-region, these sub-regions are labeled as high-frequency active areas, low-frequency active areas, or inactive areas. For example, a high-frequency threshold and a low-frequency threshold can be set; areas above the high-frequency threshold are high-frequency active areas, areas below the low-frequency threshold are inactive areas, and areas in between are low-frequency active areas. The partitioning results can generate a regional heat map.

[0071] Finally, for high-frequency activity areas, control instructions are automatically generated. These instructions can include specific action suggestions, such as "increase the density of monitoring and capture devices in area A to enhance monitoring" or "deploy insecticide in area B to reduce rodent density," along with the specific deployment coordinates. These instructions can be sent to the handheld terminals of relevant managers to guide them in implementing precise control measures.

[0072] Based on the above, individual trajectory data is transformed into operable spatial information, realizing a closed loop of monitoring and prevention, and improving the targeting and efficiency of vector-borne disease control.

[0073] It should be understood that, although Figure 2-8 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 2-8 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0074] Secondly, this embodiment provides an intelligent monitoring system for rodents, such as... Figure 9As shown, it includes: multiple monitoring and capture devices, an event acquisition module, a location mapping module, a graph construction module, an individual identification module, a trajectory fitting module, and an early warning module.

[0075] The monitoring and capture device is deployed within a pre-defined monitoring area to generate a trigger event in response to the entry of a target rodent.

[0076] The event acquisition module communicates with multiple monitoring and capture devices to acquire multiple trigger events.

[0077] The location mapping module is connected to the event acquisition module. It is used to query a preset device location mapping table based on device identification information, map the triggered event to the corresponding location on the preset monitoring area map, and generate a location marker event containing location information and trigger timestamp.

[0078] The graph construction module is connected to the location mapping module and is used to construct a spatiotemporal association graph based on multiple location marker events, with location marker events as nodes and the order of trigger timestamps as edges.

[0079] The individual identification module is connected to the graph construction module and is used to input the spatiotemporal correlation graph into the individual identification model. The individual identification model solves the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible movement speed constraint and outputs the location marker event sequence belonging to the same target rodent.

[0080] The trajectory fitting module is connected to the individual identification module. It is used to connect adjacent location marker events in the sequence of location marker events belonging to the same target rodent according to the order of trigger timestamps, and fit and generate the activity trajectory data of the target rodent.

[0081] The early warning module is connected to the trajectory fitting module and is used to generate early warning information when the activity trajectory data meets the preset early warning conditions.

[0082] In one embodiment, the individual identification module includes a speed constraint determination unit.

[0083] Specifically, the speed constraint determination unit is used to obtain a speed probability distribution function obtained through statistical learning of all triggering event pairs within a preset historical time period. This speed probability distribution function represents the probability distribution of the target rodent's movement speed under a given time interval. It is also used to determine the maximum possible movement speed corresponding to any given time interval based on the speed probability distribution function and a preset confidence level. The speed value corresponding to the maximum possible movement speed satisfies the following condition: the cumulative probability that the movement speed is less than or equal to that speed value equals the confidence level.

[0084] Furthermore, the velocity constraint determination unit includes a distribution function acquisition sub-unit.

[0085] Specifically, the distribution function acquisition subunit is used to extract all trigger event pairs within a preset historical time period from the database. Each trigger event pair includes the time interval and spatial distance corresponding to two consecutive triggers by the same monitoring and capture device or sequential triggers by different monitoring and capture devices. It is used to calculate the ratio of the spatial distance to the time interval for each trigger event pair, which serves as the instantaneous movement speed corresponding to each trigger event pair. Based on the magnitude of the time interval, it divides the instantaneous movement speed into multiple preset time interval intervals, each time interval corresponding to a continuous time range. It performs kernel density estimation on all instantaneous movement speeds within each time interval interval to obtain the velocity probability density function for that time interval. Finally, it performs parameterized fitting on the velocity probability density functions within different time interval intervals to obtain a continuous velocity probability distribution function.

[0086] In addition, the velocity constraint determination unit includes a confidence dynamic adjustment subunit.

[0087] Specifically, the confidence dynamic adjustment subunit, upon receiving a capture confirmation signal from the monitoring and capture device, associates the signal with the corresponding trigger event to obtain positive samples truly belonging to the same target rodent and negative samples truly belonging to different target rodents. It then calculates the recall and precision at the current confidence level based on the positive and negative samples. Furthermore, it lowers the confidence level when the recall is below a preset recall threshold and raises the confidence level when the precision is below a preset precision threshold.

[0088] In one embodiment, the graph building module includes an association graph building unit.

[0089] Specifically, the association graph construction unit is used to construct directed edges with each location marker event as a node, in the order of their trigger timestamps, to obtain an initial directed graph. It is used to calculate the time interval and spatial distance between any two nodes connected by a directed edge in the initial directed graph. If the spatial distance is less than or equal to the product of the maximum possible movement speed and the time interval, the directed edge between the two nodes is retained; otherwise, the directed edge is deleted. Each retained directed edge is assigned a weight, which is a monotonically decreasing function of the time interval and spatial distance, with a larger weight between nodes whose time interval is shorter and whose spatial distance is closer. Finally, the set of weighted directed edges is used as the spatiotemporal association graph.

[0090] Furthermore, the individual identification module includes an event sequence acquisition unit.

[0091] Specifically, the event sequence acquisition unit is used to search for the path partition with the largest edge weight in the spatiotemporal correlation graph using an individual identification model based on the maximum spanning tree algorithm. It also uses the location-marked event sequence corresponding to the nodes contained in each searched path as a location-marked event sequence belonging to the same target rodent.

[0092] In one embodiment, the system also includes a prevention and control module.

[0093] Specifically, the prevention and control module is used to count the frequency of rodent passage at the locations of each monitoring and capture device based on multiple activity trajectory data. It then divides the preset monitoring area into high-frequency activity zones, low-frequency activity zones, and inactive zones based on the frequency of passage. Finally, it generates prevention and control instructions for the high-frequency activity zones, including suggestions for increasing the density of monitoring and capture devices or placing insecticides.

[0094] It should be noted that the specific limitations regarding the intelligent rodent monitoring system can be found in the limitations of the intelligent rodent monitoring method described above, and will not be repeated here. The connections between the modules can be network communication connections, internal bus connections, or program interface calls; the specific implementation depends on the system architecture. For example, all modules can be integrated on the same server and interact through memory data sharing or function calls; alternatively, a microservice architecture can be adopted, with each module deployed on different computing nodes and working collaboratively through network communication. In this embodiment, the monitoring and capture device and the server are connected via a wireless network, and the modules within the server are connected using a message queue-based data flow method to ensure the system's high concurrency processing capability and scalability.

[0095] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for intelligent monitoring of rodents. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0096] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0097] Thirdly, a computer device is provided, including a memory and a processor. The memory is communicatively connected to the processor, and the memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the intelligent rodent monitoring method described above. Furthermore, the specific limitations of this computer device in implementing the intelligent rodent monitoring method can be found in the limitations of the intelligent rodent monitoring method described above, and will not be repeated here.

[0098] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the intelligent rodent monitoring method described above. Furthermore, the specific limitations of the computer-readable storage medium in implementing the intelligent rodent monitoring method can be found in the limitations of the intelligent rodent monitoring method described above, and will not be repeated here.

[0099] 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. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0100] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0101] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for intelligent monitoring of rodents, characterized in that, The method includes: Multiple trigger events are acquired. These multiple trigger events are generated by multiple monitoring and capture devices deployed in a preset monitoring area in response to the entry of a target rodent. Each trigger event includes device identification information and a trigger timestamp. Based on the device identification information, a preset device location mapping table is queried, and the triggering event is mapped to the corresponding location on a preset monitoring area map to generate a location marker event containing location information and a trigger timestamp; Based on multiple location marker events, a spatiotemporal association graph is constructed with the location marker events as nodes and the order of their trigger timestamps as edges; The spatiotemporal correlation graph is input into the individual identification model, which solves the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible movement speed constraint, and outputs the location marker event sequence belonging to the same target rodent; According to the order of the trigger timestamps, adjacent location marker events in the location marker event sequence belonging to the same target rodent are connected in sequence to fit and generate the activity trajectory data of the target rodent; When the activity trajectory data meets the preset warning conditions, a warning message is generated.

2. The method according to claim 1, characterized in that, The maximum possible moving speed constraint is determined in the following way: Obtain the velocity probability distribution function obtained by statistical learning based on all trigger events within a preset historical time period. The velocity probability distribution function represents the probability distribution of the movement speed of the target rodent under a given time interval. Based on the speed probability distribution function and a preset confidence level, the maximum possible moving speed corresponding to any given time interval is determined, wherein the speed value corresponding to the maximum possible moving speed satisfies the following condition: the cumulative probability that the moving speed is less than or equal to the speed value is equal to the confidence level.

3. The method according to claim 2, characterized in that, The step of obtaining the velocity probability distribution function based on statistical learning of all triggered event pairs within a preset historical time period includes: Extract all trigger event pairs within a preset historical time period from the database. Each trigger event pair includes the time interval and spatial distance corresponding to two consecutive triggers by the same monitoring and capture device or the sequential triggers by different monitoring and capture devices. Calculate the ratio of the spatial distance to the time interval for each of the trigger event pairs, and use it as the instantaneous movement speed corresponding to each of the trigger event pairs; Based on the magnitude of the time interval, the instantaneous movement speed is divided into multiple preset time interval intervals, each of which corresponds to a continuous time range; Kernel density estimation is performed on all instantaneous movement velocities within each time interval to obtain the velocity probability density function within that time interval; By parametrically fitting the velocity probability density function within different time intervals, a continuous velocity probability distribution function is obtained.

4. The method according to claim 2, characterized in that, The confidence level is dynamically adjusted based on captured verification feedback, and the dynamic adjustment includes: When a capture confirmation signal is received from the monitoring and capture device, the capture confirmation signal is associated with the corresponding triggering event to obtain positive samples that truly belong to the same target rodent and negative samples that truly belong to different target rodents. Based on the positive and negative samples, calculate the recall and precision at the current confidence level; When the recall rate is lower than a preset recall rate threshold, the confidence level is reduced; when the precision rate is lower than a preset precision rate threshold, the confidence level is increased.

5. The method according to claim 1, characterized in that, The construction of a spatiotemporal relationship graph based on multiple location marker events, with the location marker events as nodes and the order of their trigger timestamps as edges, includes: Using each location marker event as a node, directed edges are constructed according to the order of the trigger timestamps to obtain an initial directed graph; For any two nodes connected by a directed edge in the initial directed graph, calculate the time interval and spatial distance between the corresponding two position marker events; If the spatial distance is less than or equal to the product of the maximum possible movement speed and the time interval, retain the directed edge between the two nodes; otherwise, delete the directed edge. Each of the retained directed edges is assigned an edge weight, which is a monotonically decreasing function of time interval and spatial distance, with the edge weight between two nodes having a shorter time interval and a closer spatial distance being larger. The set of directed edges that have been weighted is used as the spatiotemporal relation graph.

6. The method according to claim 5, characterized in that, The individual identification model solves for the optimal path partitioning of the spatiotemporal correlation graph based on the maximum possible moving speed constraint, and outputs a sequence of location-marked events belonging to the same target rodent, including: The individual identification model is based on the maximum spanning tree algorithm, which searches for the path partition with the maximum edge weight in the spatiotemporal correlation graph. The sequence of location marker events corresponding to the nodes contained in each searched path is taken as the sequence of location marker events belonging to the same target rodent.

7. The method according to any one of claims 1 to 6, characterized in that, The method also includes: Based on multiple activity trajectory data, the frequency of rodents passing through the location of each monitoring and capture device is counted. Based on the frequency of passage, the preset monitoring area is divided into a high-frequency activity area, a low-frequency activity area, and an inactive area; Generate control instructions for the high-frequency activity area, including suggestions for increasing the density of the monitoring and capture devices or for placing killing agents.

8. An intelligent monitoring system for rodents, characterized in that, The system includes: Multiple monitoring and capture devices are deployed within a preset monitoring area to generate trigger events in response to the entry of a target rodent. The trigger events include device identification information and a trigger timestamp. The event acquisition module is communicatively connected to the plurality of monitoring and capturing devices and is used to acquire the plurality of triggering events; The location mapping module, connected to the event acquisition module, is used to query a preset device location mapping table based on the device identification information, map the triggering event to the corresponding location on a preset monitoring area map, and generate a location marker event containing location information and a trigger timestamp. The graph construction module, connected to the location mapping module, is used to construct a spatiotemporal association graph based on multiple location marker events, with the location marker events as nodes and the order of their trigger timestamps as edges; An individual identification module, connected to the graph construction module, is used to input the spatiotemporal association graph into the individual identification model. The individual identification model solves the optimal path partitioning of the spatiotemporal association graph based on the maximum possible movement speed constraint and outputs a sequence of location marker events belonging to the same target rodent. The trajectory fitting module, connected to the individual identification module, is used to sequentially connect adjacent position marker events in the position marker event sequence belonging to the same target rodent according to the order of the trigger timestamps, and fit and generate the activity trajectory data of the target rodent. The early warning module, connected to the trajectory fitting module, is used to generate early warning information when the activity trajectory data meets preset early warning conditions.

9. A computer device comprising a memory and a processor, the memory being communicatively connected to the processor, and the memory storing a computer program executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent rodent monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent rodent monitoring method as described in any one of claims 1 to 7.

Images