A digital management system and method for bio-safety access of a pig farm

Abstract

The present application relates to the technical field of livestock safety management, and particularly relates to a digital management system and method for biological safety access of a breeding pig farm. The system is cooperatively operated by a management platform, a mobile terminal, a video monitoring system and an Internet of Things device, controls the whole process of personnel, vehicles, materials, food and feed entering and leaving the breeding pig farm, and builds an integrated closed-loop management mechanism of "plan, execution, supervision, rectification and data". Through plan reporting and approval, the orderly arrangement of access activities is realized. Through job reporting and standard execution, the key operation process is standardized. Through video monitoring and parameter collection, the whole process is supervised. Through the negative list and rectification mechanism, a closed loop of problem processing is formed. Through data collection and report analysis, the access management information is visualized and decision support is realized. The method can effectively improve the biological safety management level, reduce the risk of disease transmission and improve the efficiency of breeding pig breeding.

Description

Technical Field

[0001] This invention relates to the field of livestock safety management technology, and in particular to a digital management system and method for biosafety entry and exit of breeding pig farms. Background Technology

[0002] With the continuous development of large-scale pig farming, biosecurity has become a crucial link in ensuring stable production and preventing the spread of diseases in breeding pig farms. The entry and exit of personnel, vehicles, and materials are important pathways for pathogen transmission. However, current biosecurity entry and exit management in breeding pig farms largely relies on manual records and paper-based approvals, with low levels of informatization. Common problems include non-standardized reporting of plans and a lack of unified standards for approval processes. In practice, key operations such as sampling, cleaning, disinfection, and drying rely on experience, lacking quantifiable standards and process records, making it difficult to ensure consistency. Meanwhile, supervision mainly relies on manual inspections, which suffers from insufficient coverage and delays, making it difficult to detect violations in a timely manner. Corrective measures lack closed-loop management, responsibilities are unclear, the effectiveness of implementation is difficult to track, and data from each stage is scattered, failing to form effective analysis and decision support. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides a digital management system and method for biosecurity access to and from pig farms.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides a digital management system for biosecurity access to and from a breeding pig farm, the system comprising: Plan Reporting Module: The plan reporting module is used to report the entry and exit behavior of breeding pig farms or production line plans related to biosafety incidents, and upload them to the management platform to execute the plan approval process and form a formal plan; Job execution module: The job execution module is responsible for receiving formal plans and initiating job reporting instructions. After the job is reported, the formal plan is executed according to the execution steps and standard parameters preset by the management platform. Video monitoring module: The video monitoring module is used to automatically start the monitoring camera to record and push the monitoring video in real time after the operation is reported and a monitoring command is issued. Logistics network monitoring module: The logistics network monitoring module is responsible for collecting key operational parameters in real time, performing qualification judgment analysis on preset standard parameters, and sending the judgment results to the management platform. Process rectification module: The process rectification module is used to stop the operation of the operation execution module and initiate rectification management instructions, send the negative list to the management platform, and receive and accept rectification supporting materials; Data Indicator Module: The data indicator module is responsible for collecting data on the entire process of biosecurity entry and exit of the breeding pig farm, and automatically summarizing it into multi-dimensional reports and uploading them to the management platform.

[0005] The second aspect of this invention provides a digital management method for biosecurity access to and from a pig farm, applied to the aforementioned digital management system for biosecurity access to and from a pig farm, specifically including the following steps: S01: By pre-setting biosafety event classification and event process types through the user login management platform, and configuring execution steps and setting standard parameters for each event process type, a data index tree is constructed to form a standard execution index library for the classification management of biosafety events in breeding pig farms; S02: Utilize the established field of view function range, deployment location, and multi-view scene image data of monitoring different pig farm organizations by each monitoring camera device to perform occlusion analysis on the field of view space of the pig farm organization and operation type jurisdiction that the monitoring camera device can observe to the maximum extent, and bind different monitoring camera devices to specific pig farm organizations and operation types for management. S03: Similarly, based on the above binding management steps, different intelligent IoT devices are bound to specific pig farm organizations and different operation types for management; S04: The plan information of the pig farm entry and exit plan is reported by the user on the mobile terminal at the scheduled time, and the pig farm entry and exit plan is transmitted to the management platform of the entry and exit management user for review. After approval, the formal plan is exported. S05: Through biosafety operation, users receive formal plans and initiate work reports on their mobile devices. After the report is completed, the operation is performed according to the preset execution steps and standard parameters on the management platform. S06: During the operation, the monitoring camera device automatically starts recording and pushes the real-time video stream to the video supervision user. At the same time, the IoT device collects the operation parameters in real time and compares them with the standard parameters to determine whether they are qualified, so as to supervise the operation process in real time. After the operation process is confirmed to be qualified through supervision, a confirmation command is initiated on the management platform to enter the next stage. S07: When video monitoring users discover violations during the monitoring process, they fill in a rectification list through the management platform and send it. The responsible user receives it through the mobile terminal and manages the rectification according to the requirements of the list. S08: The automatically collected data on the entire process of biosecurity entry and exit from the breeding pig farm is managed and fitted in parallel to form a summary and statistical analysis of different key management indicators, generate a multi-dimensional full-process management report, and upload it to the mobile terminal for display.

[0006] Preferably, step S01 specifically includes the following steps: The system acquires predefined biosafety events from different pig farms, along with one or more event process types covered by each biosafety event, and also acquires the management logic criteria for biosafety entry and exit from pig farms; wherein, the biosafety events include personnel, vehicles, materials, food, and feed; We define biosafety events as nodes without links and event flow types as directed edges. Based on management logic criteria, we perform logical dependency analysis and topology fitting between nodes without links and directed edges to obtain a directed acyclic management graph of biosafety events and event flow types. By uniquely finding one or more in-degree positions with an in-degree value of 0 in the directed acyclic graph, marking them as in-degree anchor points, and sequentially traversing the global predecessor nodes of each local in-degree anchor point according to the topological order of the directed acyclic graph, an index mapping tree with event execution direction is established. Based on the biosafety access control system, the pig farm obtains the prescribed execution steps for different event process types and the standard execution parameters set for each step. A hash algorithm is introduced, and the index path corresponding to the event process type is uniquely encoded in the hash algorithm by combining the prescribed execution steps and standard execution parameters to generate a joint index hash item. Based on the joint index hash item, reference pointers are constructed in the index mapping tree, and the reference pointers are made to point to the same predecessor node without links according to the context information of the index path. This creates a single-node reference structure for the directed edge corresponding to each event flow type, and finally generates a standard execution index library for the classification and management of biosafety events in breeding pig farms.

[0007] Preferably, step S02 specifically includes the following steps: Obtain site design diagrams of different breeding pig farms within the target area and the monitoring deployment network installed inside the breeding pig farms. Extract the actual deployment locations of monitoring camera devices with different numbers through the monitoring deployment network. Based on the equipment model information, obtain the predetermined field of view function range that different monitoring camera devices can observe and monitor during operation, and extract multi-view scene image data of each monitoring camera device observing different breeding pig farm organizations under the premise of actual deployment location under the premise of executing the predetermined field of view function range. Based on the site design diagram, a spatial depth domain is constructed for different pig farm organizational scenarios. The depth coordinate system of the actual deployment point is used as the projection reference. The perspective depth values ​​of the multi-view scene image data are projected into the spatial depth domain to perform layered occlusion analysis of perspective pixels and determine the occlusion occupancy rate of each monitoring camera device covering different spatial depth domains. If the occupancy rate is greater than the preset occupancy rate threshold, the breeding pig farm organization corresponding to the spatial depth domain of the occupancy rate is marked as a coarse-function monitoring organization, and the monitoring camera equipment with this occupancy rate is bound to the breeding pig farm organization of the corresponding coarse-function monitoring organization for management. Obtain the types of operations that can be performed and the information on the functional zoning of the corresponding breeding pig farm organization in the coarse functional monitoring organization. Based on the information on the functional zoning of the operations, construct a fine operation monitoring scale model of the coarse functional monitoring organization and divide it into sections. Output the operation monitoring jurisdiction of each operation type. Based on the jurisdictional distance values ​​of each surveillance camera device from different operational monitoring areas, a preset upsampling resolution is used to project the occupancy rate of the coarse functional monitoring organization onto the corresponding operational monitoring area and perform relative coverage analysis to obtain the coverage configuration weight value of each surveillance camera device for different operational monitoring areas. If the coverage configuration weight value is greater than the preset coverage configuration weight threshold, the operation monitoring area corresponding to the coverage configuration weight value is marked as a sub-functional monitoring area, and the monitoring camera device with this coverage configuration weight value is bound to the operation type of the corresponding sub-functional monitoring area for management.

[0008] Preferably, the step of constructing a spatial depth domain for different pig farm organizational scenarios based on the site design schematic diagram, using the depth coordinate system of the actual deployment points as the projection reference, and projecting the viewpoint depth values ​​of multi-view scene image data into the spatial depth domain for layered occlusion analysis of viewpoint pixels to determine the occlusion occupancy rate of each monitoring camera device covering different spatial depth domains, specifically includes the following steps: Based on the site design diagram, PROE modeling software was used to construct the spatial depth domain of different pig farm organizational scenarios. Fixed reference viewpoints were preset based on actual deployment sites, and the Soble operator was introduced to extract the scene pixel matrix and viewpoint depth values ​​of each multi-view scene image data. Based on the view depth value, the scene pixel matrix is ​​back-projected to the depth coordinate system of the fixed reference viewpoint in the spatial depth domain to obtain the view depth scalar point cloud of the occlusion spatial depth domain of the multi-view scene image. The view depth scalar point cloud is then reprojected back to the different view planes of the multi-view scene image data to obtain candidate occlusion view pixel points. Construct a visible field layer structure for each pixel in multi-view scene image data, and sort multiple candidate occluded view pixels at the same pixel position on different view planes in descending order according to view depth value. Based on the descending order of the results, viewpoint images of the visible field of view layer structure are rendered layer by layer to generate multi-view pseudo-occlusion images of the field of view function of each monitoring camera device during operation. The occlusion occupancy rate of each monitoring camera device in different spatial depth areas is determined by the multi-view pseudo-occlusion images.

[0009] Preferably, in step S06, the monitoring camera automatically starts recording and pushes the real-time video stream to the video monitoring user. Simultaneously, the IoT device collects operational parameters in real time and compares them with standard parameters to determine whether the operation is qualified. This specifically includes the following steps: The system collects a series of real-time operation data frames during the reporting process of the pig farm by binding target monitoring camera equipment and the Internet of Things; wherein, the real-time operation data frames include operation video frames and key operation parameters; The system synchronously acquires dynamic sensing parameters when the target surveillance camera device or Internet of Things captures each real-time operation data frame; wherein, the dynamic sensing parameters include viewing angle, color adjustment, parameter sensing interval, and sensing response rate. The LSTM spatiotemporal algorithm is used to perform long-term memory analysis on each of the real-time operation data frames to obtain the spatiotemporal observation state vector of the real-time operation process, and a static radiation network is constructed based on the implicit function of dynamic sensing parameters. By obtaining multiple sets of correct operational variables and corresponding working condition state vectors when different execution steps are operated according to the corresponding standard parameters through the operation guidelines for biosecurity entry and exit of breeding pig farms, a working condition change space field of correct operational variables is constructed based on the change effect expressed by the working condition state vector. The vector coding algorithm is used to encode the state vectors of each spatiotemporal observation, and the topological code symbol of the operating conditions in real time is output. The topological code symbol of the operating conditions is introduced into the operating condition change space field for extended analysis to obtain the dynamic hidden topological space of the real-time operating conditions. The static radiative network is reconstructed and modeled in the dynamic implicit topology space using a pre-defined standard parameter operating condition feature space. During the reconstruction modeling process, a slicing function is preset based on the supervision time sequence generated by the real-time operation data frame. Based on the slicing function, the real-time operation data frame is put back into the working condition feature space according to the slice time sequence and the features are rendered to obtain the dynamic drift rate of the real-time operation data frame relative to the standard parameters. If the dynamic drift rate is less than the preset dynamic drift rate threshold, the current operation under real-time video supervision is deemed qualified and does not need to enter the rectification management process; if the dynamic drift rate is greater than the preset dynamic drift rate threshold, the current operation under real-time video supervision is deemed unqualified, the operation is immediately stopped and the rectification management process is initiated.

[0010] Preferably, step S08 specifically includes the following steps: Acquire full-process data on biosecurity entry and exit from breeding pig farms; wherein, the full-process data includes plan reporting data, approval data, execution process data, supervision data, and rectification data; Obtain key management indicators for the entire process of breeding pig farms; among which, the key management indicators include plan deviation rate, plan execution rate, video viewing rate, percentage of operational violations, percentage of violation types, and rectification completion rate; Parallel management node channels are constructed based on key management indicators. A full-process management dictionary is established according to the benchmark evaluation criteria of key management indicators. Interactive dictionary atoms responsible for evaluating and fitting each parallel management node channel are extracted from the full-process management dictionary. Calculate the residual contribution value between each interactive dictionary atom and each full-process data, and determine the residual responsibility matrix between different interactive dictionary atoms and each full-process data based on the residual contribution value; A singular decomposition algorithm is introduced to decompose and calculate the residual responsibility matrix, outputting the orthogonal fitting singular value of each interactive dictionary atom. Based on each orthogonal fitting singular value, the full-process data evaluation fitting address of the full-process management dictionary is orthogonally matched and basis tracing is performed to obtain the sparse regression coefficient matrix. The sparse regression coefficient matrix is ​​distributed to each parallel management node channel to perform parallel fitting on the entire process data, generate local data statistical results of different key management indicators, summarize and aggregate the local data statistical results of all key management indicators, and finally generate a multi-dimensional full-process management report on biosecurity entry and exit of the breeding pig farm and upload it to the mobile terminal for display.

[0011] This invention addresses the technical deficiencies in the prior art, and its beneficial technical effects are as follows: This invention provides a digital management system and method for biosecurity entry and exit in pig farms. By constructing a full-process digital closed-loop management mechanism of "planning—execution—supervision—rectification—data," it can achieve standardized, regulated, and visualized control over the entry and exit of personnel, vehicles, materials, food, and feed, effectively overcoming the shortcomings of traditional manual management, such as disordered planning, non-standard execution, and lagging supervision. Simultaneously, by combining video surveillance and IoT devices to collect and intelligently determine key operational parameters in real time, it achieves precise supervision throughout the entire process, significantly improving the compliance and efficiency of biosecurity operations. Furthermore, through a negative list and rectification closed-loop mechanism, it clarifies the responsible parties and rectification deadlines, ensuring that problems are traceable and implementable, significantly enhancing risk management capabilities. Relying on automatic data aggregation and multi-dimensional analysis reports, managers can grasp the overall operational status in real time and make scientific decisions, effectively reducing the risk of disease transmission, implementing timely prevention strategies, reducing breeding losses, and improving the biosecurity level and economic benefits of pig farms. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0013] Figure 1 This diagram illustrates the system framework of a digital management system for biosecurity access to and from a pig farm. Figure 2 A flowchart of the first method of a digital management approach for biosecurity access to and from a pig farm is shown. Figure 3 A flowchart of a second method for digital management of biosecurity access to and from a pig farm is shown. Detailed Implementation

[0014] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0015] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0016] The first aspect of this invention provides a digital management system for biosecurity access to and from a breeding pig farm, such as... Figure 1 As shown, the system includes: Plan reporting module 101: The plan reporting module 101 is used to report the entry and exit behavior of breeding pig farms or production line plans of biosafety incidents, and upload them to the management platform to execute the plan approval process and form a formal plan; Job execution module 102: The job execution module 102 is responsible for accepting formal plans and initiating job reporting instructions. After the job is reported, the formal plan is executed according to the execution steps and standard parameters preset by the management platform. Video monitoring module 103: The video monitoring module 103 is used to automatically start the monitoring camera to record video and push the monitoring video in real time after the operation is reported and a monitoring command is issued. Logistics network monitoring module 104: The logistics network monitoring module 104 is responsible for collecting key operation parameters in real time, performing qualification judgment analysis on preset standard parameters, and sending the judgment results to the management platform. Process rectification module 105: The process rectification module 105 is used to stop the operation of the operation execution module and initiate rectification management instructions, send the negative list to the management platform, and receive and accept rectification supporting materials; Data indicator module 106: The data indicator module 106 is responsible for collecting data on the entire process of biosecurity entry and exit of the breeding pig farm, and automatically summarizing it into multi-dimensional reports and uploading them to the management platform.

[0017] The second aspect of this invention provides a digital management method for biosecurity access to and from a pig farm, applied to the aforementioned digital management system for biosecurity access to and from a pig farm, such as... Figure 2 As shown, the specific steps include: S01: By pre-setting biosafety event classification and event process types through the user login management platform, and configuring execution steps and setting standard parameters for each event process type, a data index tree is constructed to form a standard execution index library for the classification management of biosafety events in breeding pig farms; S02: Utilize the established field of view function range, deployment location, and multi-view scene image data of monitoring different pig farm organizations by each monitoring camera device to perform occlusion analysis on the field of view space of the pig farm organization and operation type jurisdiction that the monitoring camera device can observe to the maximum extent, and bind different monitoring camera devices to specific pig farm organizations and operation types for management. S03: Similarly, based on the above binding management steps, different intelligent IoT devices are bound to specific pig farm organizations and different operation types for management; S04: The plan information of the pig farm entry and exit plan is reported by the user on the mobile terminal at the scheduled time, and the pig farm entry and exit plan is transmitted to the management platform of the entry and exit management user for review. After approval, the formal plan is exported. S05: Through biosafety operation, users receive formal plans and initiate work reports on their mobile devices. After the report is completed, the operation is performed according to the preset execution steps and standard parameters on the management platform. S06: During the operation, the monitoring camera device automatically starts recording and pushes the real-time video stream to the video supervision user. At the same time, the IoT device collects the operation parameters in real time and compares them with the standard parameters to determine whether they are qualified, so as to supervise the operation process in real time. After the operation process is confirmed to be qualified through supervision, a confirmation command is initiated on the management platform to enter the next stage. S07: When video monitoring users discover violations during the monitoring process, they fill in a rectification list through the management platform and send it. The responsible user receives it through the mobile terminal and manages the rectification according to the requirements of the list. S08: The automatically collected data on the entire process of biosecurity entry and exit from the breeding pig farm is managed and fitted in parallel to form a summary and statistical analysis of different key management indicators, generate a multi-dimensional full-process management report, and upload it to the mobile terminal for display.

[0018] It should be noted that the implementation process of step S04 is as follows: The biosafety officer (entry and exit management user) of the breeding pig farm submits the material arrival plan for the following month via computer on the 25th of each month. The planned arrival time is the 5th of the following month, and the destination is the first-line breeding shed. The plan is submitted to Zhang, the level 4 biosafety specialist. Zhang reviews the plan via computer the next day, confirms that the material arrival time does not conflict with the drying room schedule and that the relevant resources are sufficient, and clicks "Approval", generating the plan number JS20240605001. The formal plan refers to the entry and exit plan that has been formally confirmed by the entry and exit management user as executable and required to be executed.

[0019] It should be noted that the implementation process of step S05 is as follows: On the 4th of next month, biosafety operator Li receives plan JS20240605001 through a mobile APP and checks the drying room equipment, disinfectant, and sampling tools in advance; at 10:00 am on the 5th of next month, the materials are transported to the entrance of the front-line drying room. Li initiates the operation report on the mobile APP, selects the operation type "materials entering the line - drying", plan number JS20240605001, and clicks "start operation". Li completes the sampling of the outer packaging of the materials and performs rinsing, disinfection and starting of the drying equipment according to the preset steps.

[0020] Preferably, step S01 specifically includes the following steps: The system acquires predefined biosafety events from different pig farms, along with one or more event process types covered by each biosafety event, and also acquires the management logic criteria for biosafety entry and exit from pig farms; wherein, the biosafety events include personnel, vehicles, materials, food, and feed; We define biosafety events as nodes without links and event flow types as directed edges. Based on management logic criteria, we perform logical dependency analysis and topology fitting between nodes without links and directed edges to obtain a directed acyclic management graph of biosafety events and event flow types. By uniquely finding one or more in-degree positions with an in-degree value of 0 in the directed acyclic graph, marking them as in-degree anchor points, and sequentially traversing the global predecessor nodes of each local in-degree anchor point according to the topological order of the directed acyclic graph, an index mapping tree with event execution direction is established. Based on the biosafety access control system, the pig farm obtains the prescribed execution steps for different event process types and the standard execution parameters set for each step. A hash algorithm is introduced, and the index path corresponding to the event process type is uniquely encoded in the hash algorithm by combining the prescribed execution steps and standard execution parameters to generate a joint index hash item. Based on the joint index hash item, reference pointers are constructed in the index mapping tree, and the reference pointers are made to point to the same predecessor node without links according to the context information of the index path. This creates a single-node reference structure for the directed edge corresponding to each event flow type, and finally generates a standard execution index library for the classification and management of biosafety events in breeding pig farms.

[0021] It should be noted that event flow types include material entry, drying, personnel entry, and vehicle disinfection. Specified execution steps include, but are not limited to, sampling, rinsing, disinfection, and drying. Standard execution parameters include, but are not limited to, disinfection duration, drying temperature, or disinfectant concentration. Typically, biosecurity entry and exit events correspond to specific event flow types, but different event flow types can be associated with more than one biosecurity entry and exit event. For example, vehicle disinfection may simultaneously require vehicles, disinfection operators, and disinfection tools and materials. Traditional data management technologies often lead to omissions or errors in data retrieval and export of preset execution steps and standard parameters sent by the management platform during operations. For instance, when executing a biosecurity event involving personnel and vehicles, executing a vehicle disinfection event might only require one execution step and standard parameter. However, if personnel, vehicles, and materials are simultaneously undergoing vehicle disinfection, multiple execution steps and standard parameters may be needed to meet the requirements. Therefore, database technology with accurate event indexing can significantly improve the accuracy of biosecurity entry and exit management in breeding pig farms. To address this, this method first employs a topological dependency modeling—a directed acyclic graph (DAG)—to examine the management relationships and logic between biosafety events and event process types. This graph clearly defines the predecessor relationships between events and event types, providing parent node dependency clues for the subsequent construction of the index tree structure. Due to the various index association possibilities mentioned above, the directed acyclic state of biosafety events and event process types resembles an index forest structure, implying complex index query derivation entry points. This undoubtedly increases the risk of output errors in execution steps and standard step parameters. Therefore, this method identifies potential in-degree anchors, where in-degree anchors are nodes without predecessors pointed to by directed acyclic nodes. In-degree anchors are suitable as the root node base of the event index tree. By traversing and accessing global predecessor nodes at these in-degree anchors, a global dependency path from event nodes to the predecessor relationship index is formed. The constructed index mapping tree allows each event node to quickly locate the execution path source of its predecessor dependencies based on the local index of the event process type, ensuring the correct pointing of the global data index for a unique event process type under multi-dimensional biosafety event classification.

[0022] It should be noted that by combining the index path that uniquely encodes the event flow type with the prescribed execution steps and standard execution parameters, the configuration of execution steps and standard parameters for different event flow types becomes more unique and unidirectional. The joint index hash item allows the execution path to be precisely indexed; even when executing a certain event flow type under multiple biosafety events, only the specific execution steps and standard parameters can be indexed. Furthermore, the reference pointers constructed based on the joint index hash item in this method can be converted into a tree-structured index mode while maintaining the directed acyclic structure of biosafety events and event flow types. Through pointers, event types are referenced by multiple parent nodes, avoiding redundant data export and improving the global data coverage of the multi-level event classification index. This method can construct a unified standard execution database with a predecessor index effect for biosafety events, event flow types, and corresponding execution steps and standard parameters configured by users on the management platform. Compared with traditional databases, this effectively improves the accuracy of data application and operation execution references, providing reliable data support for the subsequent execution process of biosafety entry and exit management in breeding pig farms.

[0023] Preferably, the S02, as Figure 3 As shown, the specific steps include: Obtain site design diagrams of different breeding pig farms within the target area and the monitoring deployment network installed inside the breeding pig farms. Extract the actual deployment locations of monitoring camera devices with different numbers through the monitoring deployment network. Based on the equipment model information, obtain the predetermined field of view function range that different monitoring camera devices can observe and monitor during operation, and extract multi-view scene image data of each monitoring camera device observing different breeding pig farm organizations under the premise of actual deployment location under the premise of executing the predetermined field of view function range. Based on the site design diagram, a spatial depth domain is constructed for different pig farm organizational scenarios. The depth coordinate system of the actual deployment point is used as the projection reference. The perspective depth values ​​of the multi-view scene image data are projected into the spatial depth domain to perform layered occlusion analysis of perspective pixels and determine the occlusion occupancy rate of each monitoring camera device covering different spatial depth domains. If the occupancy rate is greater than the preset occupancy rate threshold, the breeding pig farm organization corresponding to the spatial depth domain of the occupancy rate is marked as a coarse-function monitoring organization, and the monitoring camera equipment with this occupancy rate is bound to the breeding pig farm organization of the corresponding coarse-function monitoring organization for management. Obtain the types of operations that can be performed and the information on the functional zoning of the corresponding breeding pig farm organization in the coarse functional monitoring organization. Based on the information on the functional zoning of the operations, construct a fine operation monitoring scale model of the coarse functional monitoring organization and divide it into sections. Output the operation monitoring jurisdiction of each operation type. Based on the jurisdictional distance values ​​of each surveillance camera device from different operational monitoring areas, a preset upsampling resolution is used to project the occupancy rate of the coarse functional monitoring organization onto the corresponding operational monitoring area and perform relative coverage analysis to obtain the coverage configuration weight value of each surveillance camera device for different operational monitoring areas. If the coverage configuration weight value is greater than the preset coverage configuration weight threshold, the operation monitoring area corresponding to the coverage configuration weight value is marked as a sub-functional monitoring area, and the monitoring camera device with this coverage configuration weight value is bound to the operation type of the corresponding sub-functional monitoring area for management.

[0024] It should be noted that the management system of this invention binds monitoring camera devices to the breeding pig farm organization and different operation types for real-time supervision of the operation process and monitoring whether the operation process meets the standards, thereby ensuring the compliance of biosecurity entry and exit operations. However, traditional management techniques usually use a connection method based on proximity to bind the breeding pig farm area to monitoring cameras. Since different breeding pig farm organizations have a large number of monitoring camera devices, including internal and external ones, and their deployment locations are not uniform, if only the proximity principle is used or the binding is random, it may lead to confusion, errors, or invalid monitoring when calling monitoring cameras for specific operation areas or execution processes. This can easily result in blind spots in monitoring of operation execution or poor field of view alignment and other video supervision phenomena. To address this, this method uses a hierarchical search model to cover the monitoring field of view of the breeding pig farm organization and operation type from a coarse spatial resolution scale to a fine spatial resolution scale. The underlying logic is that different operation types are all executed inside or outside the breeding pig farm organization, that is, the monitoring details of the operation type do not leave the spatial scalar field of the breeding pig farm organization. Therefore, this method first analyzes and quickly locates the achievable occlusion positions of monitoring camera equipment from the perspective of the breeding pig farm organization at a coarse spatial resolution scale. Specifically, a 3D site structure model of different breeding pig farms is used as the spatial description space for occlusion relationships, and the actual deployment points are used as spatial observation base points for the relative positions of monitoring camera equipment. The pixel and depth values ​​of multi-view images serve as evidence of occlusion from the subjective perspective of the monitoring camera equipment. By mapping image pixels into space, the spatial occlusion relationships of each monitoring camera at different depth layers in 3D space are analyzed, revealing the degree of occlusion of different breeding pig farm areas by the limit field of view of the monitoring camera, which is quantified by the occlusion occupancy rate. If the occupancy rate is greater than a preset occupancy rate threshold, it indicates that the monitoring camera has a large occlusion area for the field of view of the breeding pig farm organization, indicating that the camera can cover the entire operation scene of the breeding pig farm organization to the maximum extent. Therefore, the two can be deeply bound together for management. This achieves a reasonable binding effect of coarse field of view function between monitoring camera equipment and breeding pig farm organization.

[0025] It should be noted that the pig farm organization includes both the pig farm itself and the production line. Operation types include, but are not limited to, personnel entry and exit, material drying, vehicle disinfection, and feed distribution. After initially binding the monitoring camera equipment to the pig farm organization within its maximum field of view, further fine-grained regional resolution binding management is needed for different operation types within the organization. To address this, this method divides the three-dimensional scene structure of the coarse-grained monitoring organization into multiple monitoring areas, i.e., operation monitoring jurisdictions, based on the actual operation zoning (operational function zoning information). Subsequently, the occlusion quantization medium (occlusion occupancy rate) obtained at the coarse spatial resolution scale is mapped to the fine spatial layer of each jurisdiction, and the upsampling scale of different operation monitoring jurisdictions is constrained based on the jurisdictional distance value. This ensures the smooth transition of cross-scale information transmission and improves the quantification accuracy and reliability of coverage weights. If the coverage configuration weight value is greater than the preset coverage configuration weight threshold, it indicates that the monitoring camera equipment has a high priority for the field of view coverage of a certain operation type area within the pig farm organization, indicating that the maximum field of view of the monitoring camera can clearly radiate to that operation type process. This method can first quickly locate and adapt different monitoring cameras for occlusion in the pig farm organization monitoring based on coarse spatial resolution, and then gradually move to fine spatial resolution to analyze the coverage priority of different operation types and perform in-depth refinement binding. This improves the rationality of monitoring binding for biosecurity entry and exit in pig farms and ensures the correct scheduling of video supervision during the operation process.

[0026] Preferably, the step of constructing a spatial depth domain for different pig farm organizational scenarios based on the site design schematic diagram, using the depth coordinate system of the actual deployment points as the projection reference, and projecting the viewpoint depth values ​​of multi-view scene image data into the spatial depth domain for layered occlusion analysis of viewpoint pixels to determine the occlusion occupancy rate of each monitoring camera device covering different spatial depth domains, specifically includes the following steps: Based on the site design diagram, PROE modeling software was used to construct the spatial depth domain of different pig farm organizational scenarios. Fixed reference viewpoints were preset based on actual deployment sites, and the Soble operator was introduced to extract the scene pixel matrix and viewpoint depth values ​​of each multi-view scene image data. Based on the view depth value, the scene pixel matrix is ​​back-projected to the depth coordinate system of the fixed reference viewpoint in the spatial depth domain to obtain the view depth scalar point cloud of the occlusion spatial depth domain of the multi-view scene image. The view depth scalar point cloud is then reprojected back to the different view planes of the multi-view scene image data to obtain candidate occlusion view pixel points. Construct a visible field layer structure for each pixel in multi-view scene image data, and sort multiple candidate occluded view pixels at the same pixel position on different view planes in descending order according to view depth value. Based on the descending order of the results, viewpoint images of the visible field of view layer structure are rendered layer by layer to generate multi-view pseudo-occlusion images of the field of view function of each monitoring camera device during operation. The occlusion occupancy rate of each monitoring camera device in different spatial depth areas is determined by the multi-view pseudo-occlusion images.

[0027] It should be noted that, regarding the spatial occlusion relationships of surveillance cameras at different depth layers in the 3D pig farm space under coarse spatial resolution, this method specifically involves modeling various 3D spatial models of pig farms or production lines using modeling software, i.e., the spatial depth domain. This spatial depth domain contains a depth scalar field representing the viewpoint model. The scene pixel matrix reflects the two-dimensional elements of the geometric scene corresponding to various pig farm organizations captured by different surveillance camera devices, covering regional textures, contours, and lines of the pig farm scene. The viewpoint depth value concretely represents the occlusion visibility of the camera's viewpoint on the bird's-eye view of the pig farm organization, defining the subjective viewpoint input tone of the camera. Next, by back-projecting the scene pixel matrix onto the depth coordinate system of a fixed reference viewpoint in the spatial depth domain, where the fixed reference viewpoint is a single-view layered representation of pixel depth resolution, alignment base points are determined for the relative observation positions of different surveillance cameras in the global space. This considers the single-view layering accuracy and fusion stability of multi-view image content within the pig farm organization space, ensuring the spatial realism and anchoring accuracy of occluded viewpoint pixels. The viewpoint depth scalar point cloud is the occlusion fusion result of multi-view scene images. Subsequently, by collecting candidate occluded pixels from different viewpoints at the same viewpoint pixel location and sorting them by depth value, each occluded pixel is no longer a single value but a hierarchical sequence of viewpoint occlusion depth based on image feedback. This abstracts the maximum scene features captured by the monitoring camera from multiple viewpoints relative to the spatial occlusion relationship of the breeding pig farm organization, forming a pseudo-image representation based on the viewpoint occlusion spatial scene pattern. This method can project the viewpoint images of the monitoring camera from its subjective perspective onto the three-dimensional scene space of the breeding pig farm organization for pixel occlusion analysis. This clarifies the effective coverage of the maximum functional field of view of the monitoring camera equipment for different breeding pig farm organizations, providing a binding decision basis for the regional operational association between each monitoring camera and different breeding pig farm organizations. This ensures that each camera can accurately capture the process details of the breeding pig farm operations, significantly improving the scheduling accuracy of video surveillance.

[0028] Preferably, in step S06, the monitoring camera automatically starts recording and pushes the real-time video stream to the video monitoring user. Simultaneously, the IoT device collects operational parameters in real time and compares them with standard parameters to determine whether the operation is qualified. This specifically includes the following steps: The system collects a series of real-time operation data frames during the reporting process of the pig farm by binding target monitoring camera equipment and the Internet of Things; wherein, the real-time operation data frames include operation video frames and key operation parameters; The system synchronously acquires dynamic sensing parameters when the target surveillance camera device or Internet of Things captures each real-time operation data frame; wherein, the dynamic sensing parameters include viewing angle, color adjustment, parameter sensing interval, and sensing response rate. The LSTM spatiotemporal algorithm is used to perform long-term memory analysis on each of the real-time operation data frames to obtain the spatiotemporal observation state vector of the real-time operation process, and a static radiation network is constructed based on the implicit function of dynamic sensing parameters. By obtaining multiple sets of correct operational variables and corresponding working condition state vectors when different execution steps are operated according to the corresponding standard parameters through the operation guidelines for biosecurity entry and exit of breeding pig farms, a working condition change space field of correct operational variables is constructed based on the change effect expressed by the working condition state vector. The vector coding algorithm is used to encode the state vectors of each spatiotemporal observation, and the topological code symbol of the operating conditions in real time is output. The topological code symbol of the operating conditions is introduced into the operating condition change space field for extended analysis to obtain the dynamic hidden topological space of the real-time operating conditions. The static radiative network is reconstructed and modeled in the dynamic implicit topology space using a pre-defined standard parameter operating condition feature space. During the reconstruction modeling process, a slicing function is preset based on the supervision time sequence generated by the real-time operation data frame. Based on the slicing function, the real-time operation data frame is put back into the working condition feature space according to the slice time sequence and the features are rendered to obtain the dynamic drift rate of the real-time operation data frame relative to the standard parameters. If the dynamic drift rate is less than the preset dynamic drift rate threshold, the current operation under real-time video supervision is deemed qualified and does not need to enter the rectification management process; if the dynamic drift rate is greater than the preset dynamic drift rate threshold, the current operation under real-time video supervision is deemed unqualified, the operation is immediately stopped and the rectification management process is initiated.

[0029] It should be noted that after the operation is executed, the monitoring camera automatically starts recording to supervise the operation process in real time, and the intelligent IoT device collects key parameters of the operation in real time, and pushes the real-time video stream and key parameters to the video specialist user for operation observation. However, due to the high-dimensional temporal sequence and uncertainty of the operation behavior data and key parameters during the operation, it is difficult to coordinate and quantify the differences between the real-time operation status and the operation condition based on video and IoT observation. This makes it impossible to intuitively compare the operation data output of real-time video and the IoT parameter output with the preset standard parameters, which undoubtedly increases the management difficulty for the video specialist user to judge the compliance of the operation. To address this, this method first constructs a static radiation network based on the dynamic sensing parameters of the monitoring camera device or IoT real-time data push. This static radiation network is essentially an implicit neural network of video spatial coordinates, video viewing angle direction and perception sampling rhythm. It can be used to transform the complex topological changes of discrete real-time operation data frames into continuous functions, thereby decoupling the behavior itself from the observation conditions. High-dimensional observation condition labels are attached to each real-time operation data frame to eliminate the judgment error introduced by the difference between standard operation parameters and supervision by different devices, and ensure the judgmentability and accuracy of operation compliance under the supervision and constraints of standard parameters. Then, by mapping the dynamic points of the pig farm operation status with the specified execution steps and corresponding standard parameters to the standard working condition space, the continuous change characteristics and correct variable effects of the standard operation are explained, i.e. the working condition change space field. This provides the working condition topology basis for the dynamic rigid modeling of the state expression in the real-time operation data frame, and describes the working condition state change trajectory of the correct operation at different time stages.

[0030] It should be noted that by performing structured state encoding on the real-time state vector and introducing the encoding into the operational condition change space field for extended analysis, a high-dimensional latent space, namely the dynamic latent topological space, is constructed. The introduction of this dynamic latent topological space allows the complex operational condition changes contained in the real-time operation data frame to be decomposed and transformed into continuous change information in a high-dimensional topology. For example, the real-time drying action of materials during operation can be output as uniform or non-uniform drying in a high-dimensional topology, thus representing the position and path of the current operation in the standard space, identifying incorrect sequences, missing steps, and abnormal operation details, and achieving accurate upper-level operational condition quantification under standardized parameter constraints of real-time operation data. By reconstructing and modeling the static radiation network in the dynamic latent topological space, the originally "fragmented" real-time topological changes can be transformed into high-dimensional continuous functions, smoothing out the operational condition topological changes of the current operation, and thus representing the true operational condition trajectory. This method can introduce complex and uncertain operational data during real-time operations into a high-dimensional latent space based on standard execution parameters for continuous high-dimensional quantization, thereby enabling real-time comparison between operational parameters and standard execution parameters during video and IoT supervision, and improving the concreteness and accuracy of operational supervision qualification assessment.

[0031] Preferably, step S08 specifically includes the following steps: Acquire full-process data on biosecurity entry and exit from breeding pig farms; wherein, the full-process data includes plan reporting data, approval data, execution process data, supervision data, and rectification data; Obtain key management indicators for the entire process of breeding pig farms; among which, the key management indicators include plan deviation rate, plan execution rate, video viewing rate, percentage of operational violations, percentage of violation types, and rectification completion rate; Parallel management node channels are constructed based on key management indicators. A full-process management dictionary is established according to the benchmark evaluation criteria of key management indicators. Interactive dictionary atoms responsible for evaluating and fitting each parallel management node channel are extracted from the full-process management dictionary. Calculate the residual contribution value between each interactive dictionary atom and each full-process data, and determine the residual responsibility matrix between different interactive dictionary atoms and each full-process data based on the residual contribution value; A singular decomposition algorithm is introduced to decompose and calculate the residual responsibility matrix, outputting the orthogonal fitting singular value of each interactive dictionary atom. Based on each orthogonal fitting singular value, the full-process data evaluation fitting address of the full-process management dictionary is orthogonally matched and basis tracing is performed to obtain the sparse regression coefficient matrix. The sparse regression coefficient matrix is ​​distributed to each parallel management node channel to perform parallel fitting on the entire process data, generate local data statistical results of different key management indicators, summarize and aggregate the local data statistical results of all key management indicators, and finally generate a multi-dimensional full-process management report on biosecurity entry and exit of the breeding pig farm and upload it to the mobile terminal for display.

[0032] It should be noted that the system automatically collects full-process data for evaluating different key management indicators. However, the evaluation of some key management indicators may require collaborative analysis with full-process data from other indicators. For example, the percentage of work violations requires evaluation in conjunction with execution process data, supervision data, and rectification data. By constructing parallel fitting nodes for key management indicators, i.e., parallel management node channels, an independent and dedicated fitting channel can be provided for the data evaluation of each key management indicator. This allows each key management indicator to be fitted independently to the target full-process data collected at different time-series nodes. The fitting actions between these indicators do not interfere with each other but maintain data connectivity and coupling. This achieves the time-series sharing effect of full-process data, avoids the information silo phenomenon of single management indicators, and improves the accuracy and reliability of multi-dimensional indicators. The key to this method's shared and circulating evaluation of end-to-end data across parallel management nodes lies in the interactive feature anchoring between the end-to-end management dictionary and interactive dictionary atoms. Interactive dictionary atoms are the smallest feature units that fit the data commonality among key management indicators based on benchmark evaluation criteria. This allows end-to-end data to be decomposed into basic influencing factors, and each dictionary atom can participate in the evaluation of multiple indicators, improving the interpretability of different end-to-end data for the evaluation of various key management indicators.

[0033] It should be noted that the residual contribution value measures the remaining value information required by the interactive dictionary atom to evaluate and interpret different full-process data. The residual responsibility matrix separates the independent contribution of the interactive dictionary atom to fitting different full-process data in each channel; in other words, it represents the degree of responsibility of the full-process data in participating in the evaluation and fitting within each channel. Finally, the singular decomposition of the residual responsibility matrix, based on the basis pursuit of the orthogonal matching of the full-process management dictionary according to the singular decomposition results, identifies the principal fitting components of each channel. It retains the multi-dimensional connectivity key for coupling and evaluating different parallel channels, namely the sparse regression coefficient matrix, allowing each process data to enter the parallel channel in which it needs to participate in the evaluation. This ensures that the management indicator data displayed in the final report is based on a comprehensive consideration of the entire process. This method enables the automatically collected full-process data to be evaluated and fitted in parallel according to different key management indicators, while simultaneously achieving shared evaluation responsibility for the full-process data during the parallel evaluation and fitting process. The results are then aggregated into multi-dimensional report outputs, providing data support for decision-making and reporting rectification.

[0034] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A digital management system for biosecurity access to and from a pig farm, characterized in that, The system includes: Plan Reporting Module: The plan reporting module is used to report the entry and exit behavior of breeding pig farms or production line plans related to biosafety incidents, and upload them to the management platform to execute the plan approval process and form a formal plan; Job execution module: The job execution module is responsible for receiving formal plans and initiating job reporting instructions. After the job is reported, the formal plan is executed according to the execution steps and standard parameters preset by the management platform. Video monitoring module: The video monitoring module is used to automatically start the monitoring camera to record and push the monitoring video in real time after the operation is reported and a monitoring command is issued. Logistics network monitoring module: The logistics network monitoring module is responsible for collecting key operational parameters in real time, performing qualification judgment analysis on preset standard parameters, and sending the judgment results to the management platform. Process rectification module: The process rectification module is used to stop the operation of the operation execution module and initiate rectification management instructions, send the negative list to the management platform, and receive and accept rectification supporting materials; Data Indicator Module: The data indicator module is responsible for collecting data on the entire process of biosecurity entry and exit of the breeding pig farm, and automatically summarizing it into multi-dimensional reports and uploading them to the management platform.

2. A digital management method for biosecurity access to and from a pig farm, applied to the digital management system for biosecurity access to and from a pig farm as described in claim 1, characterized in that, Specifically, the following steps are included: S01: By pre-setting biosafety event classification and event process types through the user login management platform, and configuring execution steps and setting standard parameters for each event process type, a data index tree is constructed to form a standard execution index library for the classification management of biosafety events in breeding pig farms; S02: Utilize the established field of view function range, deployment location, and multi-view scene image data of monitoring different pig farm organizations by each monitoring camera device to perform occlusion analysis on the field of view space of the pig farm organization and operation type jurisdiction that the monitoring camera device can observe to the maximum extent, and bind different monitoring camera devices to specific pig farm organizations and operation types for management. S03: Similarly, based on the above binding management steps, different intelligent IoT devices are bound to specific pig farm organizations and different operation types for management; S04: The plan information of the pig farm entry and exit plan is reported by the user on the mobile terminal at the scheduled time, and the pig farm entry and exit plan is transmitted to the management platform of the entry and exit management user for review. After approval, the formal plan is exported. S05: Through biosafety operation, users receive formal plans and initiate work reports on their mobile devices. After the report is completed, the operation is performed according to the preset execution steps and standard parameters on the management platform. S06: During the operation, the monitoring camera device automatically starts recording and pushes the real-time video stream to the video supervision user. At the same time, the IoT device collects the operation parameters in real time and compares them with the standard parameters to determine whether they are qualified, so as to supervise the operation process in real time. After the operation process is confirmed to be qualified through supervision, a confirmation command is initiated on the management platform to enter the next stage. S07: When video monitoring users discover violations during the monitoring process, they fill in a rectification list through the management platform and send it. The responsible user receives it through the mobile terminal and manages the rectification according to the requirements of the list. S08: The automatically collected data on the entire process of biosecurity entry and exit from the breeding pig farm is managed and fitted in parallel to form a summary and statistical analysis of different key management indicators, generate a multi-dimensional full-process management report, and upload it to the mobile terminal for display.

3. The method for digital management of biosecurity entry and exit in a pig farm according to claim 2, characterized in that, S01 specifically includes the following steps: The system acquires predefined biosafety events from different pig farms, along with one or more event process types covered by each biosafety event, and also acquires the management logic criteria for biosafety entry and exit from pig farms; wherein, the biosafety events include personnel, vehicles, materials, food, and feed; We define biosafety events as nodes without links and event flow types as directed edges. Based on management logic criteria, we perform logical dependency analysis and topology fitting between nodes without links and directed edges to obtain a directed acyclic management graph of biosafety events and event flow types. By uniquely finding one or more in-degree positions with an in-degree value of 0 in the directed acyclic graph, marking them as in-degree anchor points, and sequentially traversing the global predecessor nodes of each local in-degree anchor point according to the topological order of the directed acyclic graph, an index mapping tree with event execution direction is established. Based on the biosafety access control system, the pig farm obtains the prescribed execution steps for different event process types and the standard execution parameters set for each step. A hash algorithm is introduced, and the index path corresponding to the event process type is uniquely encoded in the hash algorithm by combining the prescribed execution steps and standard execution parameters to generate a joint index hash item. Based on the joint index hash item, reference pointers are constructed in the index mapping tree, and the reference pointers are made to point to the same predecessor node without links according to the context information of the index path. This creates a single-node reference structure for the directed edge corresponding to each event flow type, and finally generates a standard execution index library for the classification and management of biosafety events in breeding pig farms.

4. The method for digital management of biosecurity entry and exit in a pig farm according to claim 2, characterized in that, S02 specifically includes the following steps: Obtain site design diagrams of different breeding pig farms within the target area and the monitoring deployment network installed inside the breeding pig farms. Extract the actual deployment locations of monitoring camera devices with different numbers through the monitoring deployment network. Based on the equipment model information, obtain the predetermined field of view function range that different monitoring camera devices can observe and monitor during operation, and extract multi-view scene image data of each monitoring camera device observing different breeding pig farm organizations under the premise of actual deployment location under the premise of executing the predetermined field of view function range. Based on the site design diagram, a spatial depth domain is constructed for different pig farm organizational scenarios. The depth coordinate system of the actual deployment point is used as the projection reference. The perspective depth values ​​of the multi-view scene image data are projected into the spatial depth domain to perform layered occlusion analysis of perspective pixels and determine the occlusion occupancy rate of each monitoring camera device covering different spatial depth domains. If the occupancy rate is greater than the preset occupancy rate threshold, the breeding pig farm organization corresponding to the spatial depth domain of the occupancy rate is marked as a coarse-function monitoring organization, and the monitoring camera equipment with this occupancy rate is bound to the breeding pig farm organization of the corresponding coarse-function monitoring organization for management. Obtain the types of operations that can be performed and the information on the functional zoning of the corresponding breeding pig farm organization in the coarse functional monitoring organization. Based on the information on the functional zoning of the operations, construct a fine operation monitoring scale model of the coarse functional monitoring organization and divide it into sections. Output the operation monitoring jurisdiction of each operation type. Based on the jurisdictional distance values ​​of each surveillance camera device from different operational monitoring areas, a preset upsampling resolution is used to project the occupancy rate of the coarse functional monitoring organization onto the corresponding operational monitoring area and perform relative coverage analysis to obtain the coverage configuration weight value of each surveillance camera device for different operational monitoring areas. If the coverage configuration weight value is greater than the preset coverage configuration weight threshold, the operation monitoring area corresponding to the coverage configuration weight value is marked as a sub-functional monitoring area, and the monitoring camera device with this coverage configuration weight value is bound to the operation type of the corresponding sub-functional monitoring area for management.

5. The digital management method for biosecurity entry and exit of a breeding pig farm according to claim 4, characterized in that, The process involves constructing spatial depth domains for different pig farm organizational scenarios based on site design diagrams. Using the depth coordinate system of actual deployment points as the projection reference, the perspective depth values ​​of multi-view scene image data are projected onto the spatial depth domain for layered occlusion analysis of perspective pixels. This determines the occlusion occupancy rate of each monitoring camera device covering different spatial depth domains. Specifically, this includes the following steps: Based on the site design diagram, PROE modeling software was used to construct the spatial depth domain of different pig farm organizational scenarios. Fixed reference viewpoints were preset based on actual deployment sites, and the Soble operator was introduced to extract the scene pixel matrix and viewpoint depth values ​​of each multi-view scene image data. Based on the view depth value, the scene pixel matrix is ​​back-projected to the depth coordinate system of the fixed reference viewpoint in the spatial depth domain to obtain the view depth scalar point cloud of the occlusion spatial depth domain of the multi-view scene image. The view depth scalar point cloud is then reprojected back to the different view planes of the multi-view scene image data to obtain candidate occlusion view pixel points. Construct a visible field layer structure for each pixel in multi-view scene image data, and sort multiple candidate occluded view pixels at the same pixel position on different view planes in descending order according to view depth value. Based on the descending order of the results, viewpoint images of the visible field of view layer structure are rendered layer by layer to generate multi-view pseudo-occlusion images of the field of view function of each monitoring camera device during operation. The occlusion occupancy rate of each monitoring camera device in different spatial depth areas is determined by the multi-view pseudo-occlusion images.

6. The method for digital management of biosecurity entry and exit in a pig farm according to claim 2, characterized in that, In step S06, the monitoring camera automatically starts recording and pushes the real-time video stream to the video monitoring user. Simultaneously, the IoT device collects operational parameters in real time and compares them with standard parameters to determine whether the operation is qualified. This specifically includes the following steps: The system collects a series of real-time operation data frames during the reporting process of the pig farm by binding target monitoring camera equipment and the Internet of Things; wherein, the real-time operation data frames include operation video frames and key operation parameters; The system synchronously acquires dynamic sensing parameters when the target surveillance camera device or Internet of Things captures each real-time operation data frame; wherein, the dynamic sensing parameters include viewing angle, color adjustment, parameter sensing interval, and sensing response rate. The LSTM spatiotemporal algorithm is used to perform long-term memory analysis on each of the real-time operation data frames to obtain the spatiotemporal observation state vector of the real-time operation process, and a static radiation network is constructed based on the implicit function of dynamic sensing parameters. By obtaining multiple sets of correct operational variables and corresponding working condition state vectors when different execution steps are operated according to the corresponding standard parameters through the operation guidelines for biosecurity entry and exit of breeding pig farms, a working condition change space field of correct operational variables is constructed based on the change effect expressed by the working condition state vector. The vector coding algorithm is used to encode the state vectors of each spatiotemporal observation, and the topological code symbol of the operating conditions in real time is output. The topological code symbol of the operating conditions is introduced into the operating condition change space field for extended analysis to obtain the dynamic hidden topological space of the real-time operating conditions. The static radiative network is reconstructed and modeled in the dynamic implicit topology space using a pre-defined standard parameter operating condition feature space. During the reconstruction modeling process, a slicing function is preset based on the supervision time sequence generated by the real-time operation data frame. Based on the slicing function, the real-time operation data frame is put back into the working condition feature space according to the slice time sequence and the features are rendered to obtain the dynamic drift rate of the real-time operation data frame relative to the standard parameters. If the dynamic drift rate is less than the preset dynamic drift rate threshold, the current operation under real-time video supervision is deemed qualified and does not need to enter the rectification management process; if the dynamic drift rate is greater than the preset dynamic drift rate threshold, the current operation under real-time video supervision is deemed unqualified, the operation is immediately stopped and the rectification management process is initiated.

7. The method for digital management of biosecurity entry and exit in a pig farm according to claim 2, characterized in that, S08 specifically includes the following steps: Acquire full-process data on biosecurity entry and exit from breeding pig farms; wherein, the full-process data includes plan reporting data, approval data, execution process data, supervision data, and rectification data; Obtain key management indicators for the entire process of breeding pig farms; among which, the key management indicators include plan deviation rate, plan execution rate, video viewing rate, percentage of operational violations, percentage of violation types, and rectification completion rate; Parallel management node channels are constructed based on key management indicators. A full-process management dictionary is established according to the benchmark evaluation criteria of key management indicators. Interactive dictionary atoms responsible for evaluating and fitting each parallel management node channel are extracted from the full-process management dictionary. Calculate the residual contribution value between each interactive dictionary atom and each full-process data, and determine the residual responsibility matrix between different interactive dictionary atoms and each full-process data based on the residual contribution value; A singular decomposition algorithm is introduced to decompose and calculate the residual responsibility matrix, outputting the orthogonal fitting singular value of each interactive dictionary atom. Based on each orthogonal fitting singular value, the full-process data evaluation fitting address of the full-process management dictionary is orthogonally matched and basis tracing is performed to obtain the sparse regression coefficient matrix. The sparse regression coefficient matrix is ​​distributed to each parallel management node channel to perform parallel fitting on the entire process data, generate local data statistical results of different key management indicators, summarize and aggregate the local data statistical results of all key management indicators, and finally generate a multi-dimensional full-process management report on biosecurity entry and exit of the breeding pig farm and upload it to the mobile terminal for display.

Images