Intelligent file management method and system

Abstract

The present invention provides an intelligent archive management method and system, which relates to the technical field of archive management. The method includes establishing an archive location information map, initializing an archive location database and configuring a device communication protocol; an operating terminal places the archive to be put on the shelf in an entry temporary storage area after passing through an entry facial access control; matches the shelving location according to an RFID tag and generates an instruction; dispatches an archive retrieval and return robot to perform the shelving task; the operating terminal determines the archive to be taken off the shelf after inputting an archive retrieval query instruction through an intelligent interactive terminal, matches the archive off-shelf location according to the RFID tag and generates an instruction; dispatches an archive retrieval and return robot to perform the off-shelf task; the operating terminal takes the archive away from the exit handover area after passing through an exit facial access control; dispatches a robot to perform scanning and identification according to the archive inventory instruction, generates a storage location record and updates the database, and generates an inventory report, thereby significantly improving the archive access efficiency and ensuring the accuracy, security and intelligence level of archive management.

Description

Technical Field

[0001] The present invention relates to the technical field of file management, and in particular to an intelligent file management method and system. Background Art

[0002] Currently, archive management relies heavily on traditional manual management and barcode-based automated management systems. However, traditional manual management methods are often inefficient and error-prone, while barcode-based systems, while improving management efficiency, also suffer from issues such as low recognition rates and delayed data updates.

[0003] Furthermore, existing intelligent archive management systems mostly rely on fixed-position RFID readers and manually operated archive access devices, making fully automated management difficult. Human intervention is still required to process large volumes of archives, particularly during the shelving, de-shelving, and inventory processes, where manual labor accounts for a significant portion of the workload. This fails to meet the demands of efficient, intelligent archive management, and also suffers from incomplete archival information recording, insufficient access security, and inaccuracy.

[0004] Therefore, it is necessary to provide an intelligent file management method and system to solve the above technical problems. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides an intelligent file management method and system for solving the problems of low operating efficiency, incomplete information records, and insufficient access security and accuracy in existing file management systems.

[0006] The intelligent file management method provided by the present invention includes:

[0007] Based on BIM technology, a three-dimensional archive location information map of the archive warehouse was established, the archive location database was initialized, and the equipment communication protocol was configured;

[0008] After the operator verifies his / her identity through the facial recognition access control system, he / she places the files to be put on the shelves in the temporary storage area; based on the RFID file tags of the files to be put on the shelves, he / she searches the file location database for matching target file shelving locations and generates file shelving instructions; he / she receives the file shelving instructions and dispatches the file retrieval and return robot to perform the file shelving task according to the target file shelving location;

[0009] The operating terminal inputs a file retrieval query instruction through the intelligent interactive terminal to determine the file to be removed from the shelf, and searches the file location database for a matching target file removal location based on the RFID file tag of the file to be removed from the shelf and generates a file removal instruction; receives the file removal instruction, and dispatches the file retrieval and return robot to perform the file removal task based on the target file removal location; after the file removal task is completed, the operating terminal verifies its identity through the outbound facial access control and takes away the file to be removed from the outbound handover area;

[0010] According to the file inventory instruction, the file retrieval and return robot is dispatched to scan and identify files in the file warehouse using a distributed inventory strategy, generate file storage location records and update the file location database, and generate a file inventory report.

[0011] Preferably, a unique file identifier is configured for each file, and the unique file identifier is bound to the file storage location record via the RFID tag;

[0012] Among them, the unique file identifier is divided into a person file identifier and an event file identifier. The person file identifier includes the person's name, gender and ID number, and the event file identifier includes the location of the event, the timestamp of the event and the event category code.

[0013] Preferably, the step of scheduling the file retrieval and return robot to perform the file shelving task according to the target file shelving location specifically includes:

[0014] Constructing a dynamic storage optimization model, and dispatching the file retrieval robot to perform the file shelving task based on the file importance score of the file to be shelved;

[0015] The file retrieval and return robot plans the file shelving task path by improving the A* algorithm, moves to the storage temporary storage area based on the file shelving task path and grabs the file to be shelved, then moves to the target file shelving position, combines UWB positioning technology and visual recognition technology to determine the file shelving grid, and places the file to be shelved in the file shelving grid.

[0016] Preferably, the heuristic function H(n) of the improved A* algorithm is as follows:

[0017] Where H(n) represents the heuristic function of the improved A* algorithm, which is the estimated cost of the retrieval robot from the node n corresponding to the current position to the node g corresponding to the target file storage location; Represents the coordinates of node n in the three-dimensional archive location information map; Represents the coordinates of node g in the three-dimensional archive location information map; represents the Euclidean distance from node n to node g of the fetch / return robot, i.e., the straight-line distance; represents the path complexity weight coefficient, ; represents the basic weight coefficient, ; represents the adjustment coefficient, ; ObstacleDensity represents the obstacle density of the archive warehouse; MaxDensity represents the maximum obstacle density of the archive warehouse; PathDifficulty(n) represents the comprehensive difficulty evaluation value of the path where node n is located; Represents the weight coefficient of the i-th path impact factor of the path where node n is located; represents the i-th path impact factor of the path where node n is located; represents the first path impact factor of the path where node n is located, that is, the channel width impact factor; w(n) represents the real-time measured width of the channel where node n is located; Indicates the minimum aisle width of the archive warehouse; Indicates the maximum aisle width of the archive warehouse; It represents the second path impact factor of the path where node n is located, that is, the obstacle impact factor; represents the Euclidean distance from node n to the ith obstacle; m represents the total number of obstacles within a radius of 5 meters centered on node n; represents the attenuation coefficient, ; Indicates the third path influence factor of the path where node n is located, that is, the illumination influence factor; represents the minimum working light intensity of the visual system of the pick-up and return robot; I(n) represents the light intensity at node n; represents a constant, ; The fourth path influencing factor of the path where node n is located, that is, the compact rack motion state factor; Represents the speed influence coefficient; v(n) represents the movement speed of the compact shelving in the channel where node n is located.

[0018] Preferably, the center coordinates of the file shelf determined by combining UWB positioning technology and visual recognition technology for:

[0019] Where, Indicates the center coordinates of the archive shelving grid in the three-dimensional archive storage location information map; Represents the confidence weight of UWB positioning; Represents the confidence weight of visual recognition; Indicates the estimated coordinates of the archive shelving grid in the 3D archive location information map provided by UWB positioning technology, obtained through triangulation by the UWB positioning base stations deployed in the archive warehouse; Indicates the estimated coordinates of the archive shelving openings in the 3D archive location information map, extracted by visual recognition technology. This is calculated by using the camera on the retrieval and return robot to identify the opening visual identifier and then performing perspective visual transformation. It represents the standard deviation of UWB positioning error, which is determined by the deployment density of UWB positioning base stations and the environmental multipath effect; represents a constant, ConfidenceScore represents the confidence level of the grid recognition output by the visual recognition technology; MaxScore represents the maximum possible confidence level of the visual recognition technology; represents the standard deviation of the visual positioning error; r represents the distance from the camera on the retrieval and return robot to the file storage opening; represents the camera pixel error; c represents the camera focal length.

[0020] Preferably, the position verification threshold of the file shelf is calculated , where k represents the position safety factor;

[0021] when When , it is necessary to determine the center coordinates of the file shelf opening again, where Indicates the reference coordinates of the archive shelving grid stored in the grid database.

[0022] Preferably, the receiving of the file removal instruction and scheduling the file retrieval and return robot to perform the file removal task according to the target file removal location specifically includes:

[0023] After receiving the file removal instruction, the file retrieval robot with sufficient power and closest to the target file removal location is preferentially dispatched to perform the file removal task;

[0024] The file retrieval and return robot plans the file unloading task path through the improved A* algorithm, moves to the target file unloading position based on the file unloading task path, adjusts the gripping force of the robotic arm through the force control algorithm, and determines the target file unloading position in combination with the binocular vision mechanism. The robotic arm takes down the file to be unloaded and moves it to the outbound handover area.

[0025] Preferably, the method further comprises: monitoring the air quality parameters of the archive warehouse in real time based on environmental control equipment; when the air quality parameters exceed a preset safety range, automatically adjusting the air quality parameters to the preset safety range through a fuzzy PID algorithm;

[0026] Based on surveillance cameras, a multimodal anomaly detection model is used to analyze the behavior of the operator in real time. If abnormal behavior is identified, an audible and visual alarm is issued and the entry and exit facial access control systems are locked;

[0027] Based on electronic archive scanning equipment, OCR and NLP technologies are used to automatically extract the archive text content in the archive warehouse, compare multiple archive text contents through data consistency verification algorithms, correct the contradictory archive text contents, and scan the archive text contents in real time through sensitive information recognition models. If sensitive information is found, the operation end is immediately notified.

[0028] Preferably, the comparing of the plurality of archive text contents by using a data consistency check algorithm and correcting the conflicting archive text contents specifically includes:

[0029] Constructing an archival knowledge graph containing all the archival text contents based on a graph neural network, wherein each archival text content serves as a node of the archival knowledge graph, and the relationship between two archival text contents serves as an edge of the archival knowledge graph;

[0030] Node embedding vectors are learned through graph neural networks, and semantically related nodes are close in the vector space. The anomaly detection model based on the graph neural network calculates the node similarity of the two node embedding vectors. If the node similarity is lower than the preset similarity threshold, there is a potential contradiction between the two nodes, and a rule-based secondary verification of the two nodes is required. When a contradiction is detected again, a revised version of the archive content is selected based on the node credibility and a corresponding archive content correction suggestion is generated;

[0031] Among them, the loss function L of the anomaly detection model of the graph neural network is:

[0032] Where L represents the loss function of the anomaly detection model of the graph neural network; V represents the node set of the archive knowledge graph; represents a node in the node set V; E represents the edge set of the archive knowledge graph; represents an edge in the edge set E; Represents nodes respectively The node embedding vector of Represents two node embedding vectors Node similarity; represents cosine similarity; Represents the node embedding vector The transpose of Represents the node embedding vector The model; Represents the node embedding vector The model; represents the sigmoid activation function; W represents the learnable weight matrix; b represents the bias term; Represents the node embedding vector and splicing operation; Represents an edge The true label of , if the content of nodes u and v is consistent, then , if the contents of nodes u and v are contradictory, then ; represents the L2 regularization term; represents the regularization strength parameter, ; Represents the set of all learnable parameters in a graph neural network;

[0033] The transfer function of each layer of the graph neural network is:

[0034] Where, Represents the node embedding vector of node u in the l+1 layer; Represents the node embedding vector of node u in layer l; represents the learnable weight matrix of layer l; represents the set of neighbor nodes of node u, ; represents the symmetric normalization constant, , represents the degree of node u, that is, the number of neighbor nodes, Represents the degree of node v, that is, the number of neighbor nodes.

[0035] Intelligent file management system, the management system includes:

[0036] Intelligent RFID electric compact cabinet, used for storing files, including compact shelves, drive motor, RFID reader and positioning sensor;

[0037] The retrieval robot is used to perform file shelving and de-shelving tasks, including a mobile chassis, a robotic arm, a vision system, an RFID reader, and a battery management system;

[0038] Facial access control equipment, used for personnel identity verification, including cameras, facial recognition devices and access control controllers;

[0039] Environmental control equipment, used to adjust the air quality parameters of the archive warehouse, including temperature and humidity sensors, air conditioners, dehumidifiers and ventilation systems;

[0040] Monitoring equipment for security monitoring of the archive warehouse, including cameras, video analysis servers, and sound and light alarms;

[0041] Electronic archival scanning equipment, used for digital scanning of archival text content, including a high-speed scanner, image processing module, and OCR recognition module;

[0042] Intelligent interactive terminal, used for interaction between the operating terminal and the intelligent archive management system, including touch screen, input device and display device;

[0043] Database server, used to store archival information, equipment status and operation records.

[0044] Compared with related technologies, the intelligent archive management method and system provided by the present invention have the following beneficial effects:

[0045] The present invention can establish a three-dimensional archive location information map of the archive warehouse based on BIM technology, initialize the archive location database and configure the equipment communication protocol; after the operation end verifies the identity through the entry face access control, the archive to be put on the shelf is placed in the entry temporary storage area; according to the RFID archive tag of the archive to be put on the shelf, the matching target archive shelving location is retrieved in the archive location database and an archive shelving instruction is generated; the archive shelving instruction is received, and the retrieval and return robot is dispatched to perform the archive shelving task according to the target archive shelving location; after the operation end inputs the retrieval query instruction through the intelligent interactive terminal, the archive to be taken off the shelf is determined, and according to the RFID archive tag of the archive to be taken off the shelf, the matching target archive off the shelf location is retrieved in the archive location database and an archive off the shelf instruction is generated; the archive off the shelf instruction is received, and the retrieval and return robot is dispatched to perform the archive off the shelf task according to the target archive off the shelf location; after the archive off the shelf task is completed, the operation end verifies the identity through the exit face access control and takes away the archive to be taken off the shelf in the exit handover area; according to the archive inventory instruction, the retrieval and return robot is dispatched to adopt distributed inventory in the archive warehouse The strategy is used to scan and identify archives, generate archive storage location records and update the archive location database, generate archive inventory reports, and monitor the air quality parameters of the archive warehouse in real time based on environmental control equipment. When the air quality parameters exceed the preset safety range, the fuzzy PID algorithm is used to automatically adjust the air quality parameters to the preset safety range; based on surveillance cameras, a multimodal anomaly detection model is used to analyze the behavior of the operator in real time. If abnormal behavior is identified, an audible and visual alarm is issued and the entry and exit facial access control systems are locked; based on electronic archive scanning equipment, OCR and NLP technologies are used to automatically extract the archive text content of the archive warehouse, compare multiple archive text contents through a data consistency verification algorithm, correct conflicting archive text contents, and scan the archive text content in real time through a sensitive information recognition model. If sensitive information is found, the operator is immediately notified, thereby solving the problems of multiple manual operations, low efficiency, and incomplete information management in the existing technology, significantly improving the efficiency of archive access, and ensuring the accuracy, security and intelligence level of archive management.

[0046] The present invention uses an intelligent archive management system composed of intelligent RFID electric compact cabinets, archive retrieval and return robots, etc., which realizes the automation of archive management, improves archive management efficiency, reduces the risk of manual operation, and significantly improves archive access efficiency compared to existing technologies. The present invention can integrate UWB positioning technology and visual recognition technology to accurately determine the coordinates and status of storage slots, effectively improve the accuracy and security of archive access, and reduce the risk of archive loss or misplacement. The present invention can realize the electronic management and real-time updating of archive information, comprehensively record the archive life cycle, improve the intelligence level of archive management, and meet the needs of high-efficiency and intelligent archive management. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a flow chart of the intelligent file management method of the present invention;

[0048] Figure 2 A schematic diagram of the gear retrieval and return robot of the present invention;

[0049] Figure 3 This is a system block diagram of the intelligent file management system of the present invention;

[0050] Figure 4 Schematic diagram of the intelligent RFID electric compact cabinet of the present invention;

[0051] Figure 5 A schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0053] like Figure 1 FIG. 1 is a flow chart of an intelligent file management method according to an embodiment of the present invention. Figure 1The execution subject of the method shown may be a software and / or hardware device. The execution subject of the present application may include but is not limited to at least one of the following: user equipment, network equipment, etc. Among them, user equipment may include but is not limited to computers, smart phones, personal digital assistants (PDAs) and the electronic devices mentioned above. Network equipment may include but is not limited to a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers, wherein cloud computing is a type of distributed computing, a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. It includes steps S1 to S4, as follows:

[0054] S1, based on BIM technology, establishes a three-dimensional archive location information map of the archive warehouse, initializes the archive location database and configures the equipment communication protocol;

[0055] Among them, BIM (Building Information Modeling) technology is a building life cycle management technology based on a three-dimensional digital model. It constructs a digital three-dimensional space model of the archive warehouse by integrating geometric information, material properties and spatial relationships. The three-dimensional archive location information map is a visual model of the archive storage location based on three-dimensional spatial coordinates. It can integrate spatial information such as location number, compact shelving distribution, channel planning, etc. to form a digital mapping of the archive storage location. The archive location database refers to a structured database used to store the physical location information of archives. It contains metadata such as archive numbers, location coordinates, storage device associations, etc., and supports fast retrieval and location updates. The device communication protocol refers to the data transmission rules between compact shelving, retrieval and return robots, sensors and other equipment in the intelligent archive management system to ensure the standardization and compatibility of instructions and data interactions between devices. Among them, the schematic diagram of the retrieval and return robot is as follows Figure 2 shown.

[0056] Specifically, building information modeling (BIM) technology can be used to create a three-dimensional spatial model of the archive warehouse, constructing a digital mapping model that includes the shelving layout, aisle planning, and storage location coordinates. This is known as a three-dimensional archive location information map. This map integrates the warehouse's geometric dimensions, load-bearing structure, and the spatial relationships between storage devices to form a visual matrix of archive storage locations. The three-dimensional archive location information map can then be used to synchronously initialize an archive location database. This database uses a structured storage method and integrates metadata such as archive entity numbers, physical coordinates, and storage device relationships, providing data support for archive location.

[0057] In addition, it is also necessary to complete the configuration of the equipment communication protocol, establish a standardized command interaction system by setting the data transmission rules of hardware equipment such as intelligent compact shelving, retrieval and return robots, and environmental sensors, and ensure the compatibility and real-time performance of collaborative operations of multiple devices.

[0058] S2: After the operator verifies his / her identity through the facial recognition access control system, he / she places the files to be put on the shelves in the temporary storage area; based on the RFID file tag of the files to be put on the shelves, he / she searches the file location database for matching target file shelving locations and generates file shelving instructions; receives the file shelving instructions, and dispatches the file retrieval and return robot to perform the file shelving task according to the target file shelving location;

[0059] It should be noted that the "operator end" refers to the terminal corresponding to the operator, used to initiate file management commands and monitor system status. Facial access control for entry refers to an access control device based on biometric recognition technology. It collects facial features and compares them with pre-stored information to automatically verify personnel access permissions. Archives awaiting storage refer to archives that need to be shelved for storage. RFID (Radio Frequency Identification) archive tags are electronic tags attached to archived archives. They automatically identify the archive through radio frequency signals, store metadata such as the archive number and type, and support contactless data reading. The target archive shelving location refers to the optimal shelving location for the retrieval and return robot within the three-dimensional archive location information map. Archive shelving instructions are digital operation instructions generated by the system, containing parameters such as the archive identifier, target location coordinates, and the robot's execution action. They are used to drive the retrieval and return robot to complete the shelving operation. Retrieval and return robots are automated devices capable of autonomous movement, file grabbing, and intelligent scheduling. They can shelve, de-shelve, and perform inventory tasks according to instructions. They typically integrate navigation, positioning, robotic arms, and recognition modules.

[0060] After the operator completes identity verification through the facial access control system deployed at the warehouse entrance, the file to be shelved is placed in the temporary storage area for storage. The system reads the unique identification information of the file through the RFID tag attached to the file entity and performs a search operation in the file location database. Based on the file type, frequency of use and warehouse storage strategy, the optimal storage location is matched, that is, the target file shelving location, and a digital instruction containing the file identification, storage location coordinates and operating parameters is generated, that is, the file shelving instruction. After receiving the instruction, the system dispatches a file retrieval and return robot with autonomous navigation function according to the target location coordinates. The robot integrates a robotic arm grasping device and a positioning module, moves to the temporary storage area according to the preset path, completes the file grabbing and accurately stores it in the target storage location, and realizes the fully automated execution of the shelving operation.

[0061] S3: The operating terminal inputs a file retrieval query instruction through the intelligent interactive terminal to determine the file to be removed from the shelf. According to the RFID file tag of the file to be removed from the shelf, the operating terminal searches the file location database for a matching target file removal location and generates a file removal instruction. The file removal instruction is received, and the file retrieval and return robot is dispatched to perform the file removal task according to the target file removal location. After the file removal task is completed, the operating terminal verifies its identity through the outbound facial access control and takes away the file to be removed from the outbound handover area.

[0062] It can be understood that an intelligent interactive terminal refers to a human-computer interaction device that integrates display, input and communication functions, supports natural language query, command input and visual display of archival information, and is usually deployed on the archive management console. The file retrieval query instruction refers to the archive retrieval request initiated by the user through the intelligent interactive terminal, which contains query conditions such as keywords and file numbers, and is used to locate the storage location of the target file. The file de-shelf instruction corresponds to the shelving instruction, which is used to instruct the robot to take out the file from the specified storage location, including the de-shelf location, file identification and handover process parameters. The outbound handover area refers to the physical area set up in the archive warehouse, which is used to complete the physical handover of archives from the warehouse to the user, and is usually equipped with identity authentication and outbound registration equipment.

[0063] The operator submits a file retrieval query request containing keywords or file numbers through an intelligent interactive terminal with integrated touch display and command input functions. After the system parses the request, it locates the physical storage location of the file to be taken off the shelf, matches the target file off-shelf location in the database based on the RFID tag information, and generates a file off-shelf instruction containing the off-shelf coordinates and handover parameters. After receiving the instruction, the retrieval and return robot moves to the target storage location according to the optimal path planning, confirms the file information through visual recognition or RFID scanning, performs the off-shelf operation and transports it to the outbound handover area. At this time, the operator needs to verify his identity again through the facial access control and complete the file collection in the set physical handover area. The system will simultaneously update the file status database to form a closed loop of outbound records.

[0064] S4, according to the file inventory instruction, the file retrieval and return robot is dispatched to scan and identify files in the file warehouse using a distributed inventory strategy, generate file storage location records and update the file location database, and generate a file inventory report.

[0065] Among them, archive inventory instructions refer to inventory verification instructions triggered by the system or manually, which are used to initiate a systematic verification process for the location and quantity of archives to ensure that the physical archives are consistent with the database records. The distributed inventory strategy refers to dividing the archive warehouse into multiple sub-areas and performing inventory tasks through multiple retrieval and return robots working in parallel. Compared with centralized inventory methods, this can significantly improve inventory efficiency and reduce equipment resource conflicts. The archive inventory report is a structured document generated after the inventory operation is completed. It contains information such as archive quantity statistics, location deviation records, and abnormal archive lists, providing data support for archive management decisions.

[0066] Upon receiving an archive inventory command, the system initiates a distributed inventory strategy. This strategy divides the warehouse into multiple sub-areas and dispatches multiple retrieval and return robots to operate in parallel. Using their onboard RFID readers or visual recognition modules, they batch scan and verify the identification information of archive entities, generate archive storage location records, and automatically update the data in the archive location database. Finally, the system generates a structured archive inventory report, including archive quantity statistics, a list of abnormal locations, and recommended actions. This provides data support for archive management decisions and ensures consistency between physical archives and digital records.

[0067] In a specific implementation process, a unique file identifier is configured for each file, and the unique file identifier is bound to the file storage location record through the RFID tag;

[0068] Among them, the unique file identifier is divided into a person file identifier and an event file identifier. The person file identifier includes the person's name, gender and ID number, and the event file identifier includes the location of the event, the timestamp of the event and the event category code.

[0069] It's understandable that each archive can be assigned a unique archive identifier, and using RFID tags, this identifier can be associated and bound to the metadata record of the archive's physical storage location. This binding mechanism achieves a two-way mapping of archive identity information and spatial location through database indexing, forming a traceable digital management chain.

[0070] Unique archive identifiers are divided into two coding systems based on the attributes of the archive content: person archive identifiers and event archive identifiers. Person archive identifiers utilize a hierarchical data structure, encompassing identification information such as a natural person's name, gender, and citizen ID number, thereby ensuring unique identification and identity verification of individual archives. Event archive identifiers construct a coding framework based on event elements, encompassing the geographic coordinates of the event, timestamps, and event category codes. Timestamps utilize a standardized time recording format, while event category codes utilize a tree-like hierarchical system, enabling categorized management and rapid retrieval of event archives of varying natures.

[0071] Through the above-mentioned identification system and binding mechanism, the intelligent archive management system can use the non-contact recognition function of RFID tags to read archive identification in real time and associate it with its storage location, providing a standardized data foundation for the automated access, inventory verification and full life cycle management of archives.

[0072] The step of scheduling the file retrieval and return robot to perform the file shelving task according to the target file shelving location specifically includes:

[0073] Constructing a dynamic storage optimization model, and dispatching the file retrieval robot to perform the file shelving task based on the file importance score of the file to be shelved;

[0074] The file retrieval and return robot plans the file shelving task path by improving the A* algorithm, moves to the storage temporary storage area based on the file shelving task path and grabs the file to be shelved, then moves to the target file shelving position, combines UWB positioning technology and visual recognition technology to determine the file shelving grid, and places the file to be shelved in the file shelving grid.

[0075] The calculation formula for the importance score of the archives to be put on the shelf is as follows:

[0076] In the formula, DAPF represents the archive importance score of the archive to be put on the shelf; are the weight coefficients of frequency of listing, importance level and age respectively; SJPL and Respectively represent the frequency of the files to be put on the shelf and the maximum frequency of all files put on the shelf; ZYXDJ and Respectively represent the importance level of the files to be put on the shelf and the maximum value of the importance level of the files; NL and They represent the age of the file to be put on the shelf and the maximum age of all files respectively;

[0077] It should be noted that the higher the file importance score, the higher the file's priority when it comes to shelving tasks. The average number of times a file has been shelved over the past three years is the shelving frequency. The maximum importance level is fixed at 5, that is, the importance level is fixedly divided into 5 levels, among which Level 5 is the highest level of importance and corresponds to archives of core business documents and confidential information; Level 1 is the lowest level of importance and corresponds to archives of temporary reference materials; the remaining levels correspond to ordinary business archives. The interval between the creation date of the file and the current system time is the age.

[0078] And the weight coefficient of the frequency of listing Dynamic adjustment is required, and the weight coefficient of the adjusted listing frequency ,in, Represents the dynamic adjustment coefficient; DQSJL represents the current shelf quantity; LSSJL represents the historical average shelf quantity.

[0079] When executing a file shelving task, the system first needs to build a dynamic storage optimization model. This model intelligently plans storage locations based on file management strategies, warehouse space utilization efficiency, and the importance scores of the files to be shelved. The file importance scoring system comprehensively considers the multi-dimensional attributes of files to form a quantitative priority assessment, providing a basis for decision-making on storage allocation.

[0080] When dispatching the retrieval and return robots, the system uses a modified A* algorithm. This algorithm, based on a classic path search algorithm, incorporates environmental parameters such as warehouse aisle layout and equipment operating status to generate an optimal movement path from the current location to the temporary storage area and then to the target storage location. The retrieval and return robots follow this path to the temporary storage area, where they retrieve the files using their integrated gripper and then proceed along the planned path to the target storage area.

[0081] Furthermore, the retrieval robot utilizes UWB (ultra-wideband) positioning technology and visual recognition technology to determine the storage slot for the files. UWB positioning technology uses radio frequency signals to achieve centimeter-level spatial coordinate positioning, providing the robot with a macro-scale positional reference. Visual recognition technology uses image acquisition and pattern recognition algorithms to accurately identify the characteristics of the storage slots at the micro level. By fusing these two data, the retrieval robot can accurately determine the specific slot for the file to be shelved and, using its end effector, smoothly place the file in the target slot, completing the fully automated shelving process.

[0082] The heuristic function H(n) of the improved A* algorithm is as follows:

[0083] Where H(n) represents the heuristic function of the improved A* algorithm, which is the estimated cost of the retrieval robot from the node n corresponding to the current position to the node g corresponding to the target file storage location; Represents the coordinates of node n in the three-dimensional archive location information map; Represents the coordinates of node g in the three-dimensional archive location information map; represents the Euclidean distance from node n to node g of the fetch / return robot, i.e., the straight-line distance; represents the path complexity weight coefficient, ; represents the basic weight coefficient, ; represents the adjustment coefficient, ; ObstacleDensity represents the obstacle density of the archive warehouse; MaxDensity represents the maximum obstacle density of the archive warehouse; PathDifficulty(n) represents the comprehensive difficulty evaluation value of the path where node n is located; Represents the weight coefficient of the i-th path impact factor of the path where node n is located; represents the i-th path impact factor of the path where node n is located; represents the first path impact factor of the path where node n is located, that is, the channel width impact factor; w(n) represents the real-time measured width of the channel where node n is located; Indicates the minimum aisle width of the archive warehouse; Indicates the maximum aisle width of the archive warehouse; It represents the second path impact factor of the path where node n is located, that is, the obstacle impact factor; represents the Euclidean distance from node n to the ith obstacle; m represents the total number of obstacles within a radius of 5 meters centered on node n; represents the attenuation coefficient, ; Indicates the third path influence factor of the path where node n is located, that is, the illumination influence factor; represents the minimum working light intensity of the visual system of the pick-up and return robot; I(n) represents the light intensity at node n; represents a constant, ; The fourth path influencing factor of the path where node n is located, that is, the compact rack motion state factor; Represents the speed influence coefficient; v(n) represents the movement speed of the compact shelving in the channel where node n is located.

[0084] The retrieval robot's path planning utilizes a modified A* algorithm. Its core heuristic function estimates the cost of moving from the current position to the target location. This function consists of two components: the three-dimensional straight-line distance and a path complexity correction term. The three-dimensional straight-line distance is calculated based on the coordinates of the current and target nodes in the three-dimensional archive map and reflects the ideal cost of moving in the absence of obstacles. The path complexity weighting factor is dynamically adjusted using a base weight and an adjustment factor. Its value is related to the obstacle density in the archive. As the obstacle density in the environment increases, the factor automatically increases to strengthen the penalty for complex paths.

[0085] The comprehensive path difficulty assessment is composed of four weighted influencing factors: Channel width influencing factor: Based on the relative relationship between the current channel real-time width and the minimum and maximum channel widths of the warehouse, the exponential function is used to calculate the impact of the channel narrowness on the path. The closer the channel is to the minimum width, the greater the impact factor; Obstacle influencing factor: With the current node as the center, the distance attenuation effect of all obstacles within the radius is calculated. The closer the obstacle is, the greater the negative impact on the path. Quantitative evaluation is achieved through the exponential operation of the inverse of the square of the distance and the attenuation coefficient; Light influencing factor: Based on the ratio of the light intensity at the current position to the minimum working light intensity of the robot vision system, the impact of insufficient light on positioning accuracy is calculated. When the light is lower than the threshold, this factor significantly increases the path difficulty; Compact shelving motion state factor: Distinguish between the stationary and moving states of the compact shelving. In the moving state, the path difficulty increases linearly according to its speed. The higher the speed, the greater the impact factor to avoid collision between the robot and the moving compact shelving. The sum of the weight coefficients of the four factors is 1. Through multi-dimensional modeling of channel width, obstacle distribution, lighting conditions and equipment motion status, accurate estimation of path costs in complex warehouse environments is achieved, ensuring that robots plan optimal movement paths in dynamic scenarios while taking into account both efficiency and safety.

[0086] The center coordinates of the file shelf determined by combining UWB positioning technology and visual recognition technology for:

[0087] Where, Indicates the center coordinates of the archive shelving grid in the three-dimensional archive storage location information map; Represents the confidence weight of UWB positioning; Represents the confidence weight of visual recognition; Indicates the estimated coordinates of the archive shelving grid in the 3D archive location information map provided by UWB positioning technology, obtained through triangulation by the UWB positioning base stations deployed in the archive warehouse; Indicates the estimated coordinates of the archive shelving openings in the 3D archive location information map, extracted by visual recognition technology. This is calculated by using the camera on the retrieval and return robot to identify the opening visual identifier and then performing perspective visual transformation. It represents the standard deviation of UWB positioning error, which is determined by the deployment density of UWB positioning base stations and the environmental multipath effect; represents a constant, ConfidenceScore represents the confidence level of the grid recognition output by the visual recognition technology; MaxScore represents the maximum possible confidence level of the visual recognition technology; represents the standard deviation of the visual positioning error; r represents the distance from the camera on the retrieval and return robot to the file storage opening; represents the camera pixel error; c represents the camera focal length.

[0088] Calculate the location verification threshold of the file shelf , where k represents the position safety factor;

[0089] when When , it is necessary to determine the center coordinates of the file shelf opening again, where Indicates the reference coordinates of the archive shelving grid stored in the grid database.

[0090] During file shelving operations, the retrieval and return robot uses a fusion algorithm combining UWB positioning and visual recognition technology to determine the three-dimensional coordinates of the shelving slot. This fusion positioning method uses a weighted average strategy, fusing the coordinate estimates from the two technologies based on confidence weights to form the final coordinates of the slot center.

[0091] Specifically, UWB positioning technology uses triangulation measurements from positioning base stations deployed within the warehouse to provide an initial coordinate estimate of the grid opening. Its confidence weight is calculated based on the standard deviation of the positioning error, which is affected by the density of base station deployment and the multipath effect of the environment. The smaller the error, the higher the weight. Visual recognition technology uses a camera mounted on a retrieval robot to identify the visual identifier of the grid opening, such as a QR code or a characteristic pattern. The grid opening coordinates are calculated using a perspective transformation algorithm. Its confidence weight is determined by the ratio of the confidence score output by the recognition model to the maximum possible confidence score, combined with the standard deviation of the visual positioning error. The visual positioning error is related to the distance from the camera to the grid opening, pixel error, and focal length. The closer the distance or the longer the focal length, the smaller the error.

[0092] The fusion algorithm performs a weighted summation of the coordinate estimates from the two techniques and normalizes them using the sum of the confidence weights to generate the final three-dimensional coordinates of the grid opening. To ensure positioning accuracy, the system sets a position verification threshold, determined by multiplying the safety factor by the sum of the confidence weights of the two techniques. If the fused coordinates deviate from the reference coordinates stored in the database by more than this threshold, the system triggers a secondary positioning process, redundancy measures such as LiDAR scanning to re-determine the grid opening position, thereby ensuring accurate file placement.

[0093] This fusion positioning technology dynamically adjusts the weights of two types of heterogeneous positioning technologies, effectively combining the macro-positioning advantages of UWB and the micro-identification capabilities of visual recognition, achieving centimeter-level positioning accuracy in complex archive warehouse environments, and providing technical support for the automated and precise shelving of archives.

[0094] The receiving of the file removal instruction and scheduling the file retrieval and return robot to perform the file removal task according to the target file removal location specifically includes:

[0095] After receiving the file removal instruction, the file retrieval robot with sufficient power and closest to the target file removal location is preferentially dispatched to perform the file removal task;

[0096] The file retrieval and return robot plans the file unloading task path through the improved A* algorithm, moves to the target file unloading position based on the file unloading task path, adjusts the gripping force of the robotic arm through the force control algorithm, and determines the target file unloading position in combination with the binocular vision mechanism. The robotic arm takes down the file to be unloaded and moves it to the outbound handover area.

[0097] When performing file removal tasks, the system follows an intelligent scheduling strategy. Upon receiving a file removal instruction, based on multi-objective optimization principles, it prioritizes the retrieval robot with a remaining battery level above a preset threshold and the shortest Euclidean distance to the target file removal location, maximizing task execution efficiency and minimizing energy consumption. This scheduling mechanism collects real-time data on the robot's battery status and spatial location, constructing a dynamic scheduling decision model to achieve optimal allocation of equipment resources.

[0098] After the retrieval and return robot is activated, it uses an improved A* algorithm that incorporates environmental parameters, taking into account factors such as the warehouse's compact shelving aisle layout, obstacle distribution, and equipment operating status, to generate an optimal movement path from its current position to the target unloading location. When the retrieval and return robot reaches the target storage area along this path, it activates a force control strategy to adjust the gripping force of the robotic arm. This strategy, based on the estimated file weight and grasping speed feedback, dynamically adjusts the gripping force of the end effector through a closed-loop control mechanism to ensure that the file is retrieved without damage.

[0099] When locating the target file, the retrieval and return robot relies on a binocular vision imaging system, using a stereo vision matching algorithm to calculate the file's three-dimensional spatial coordinates. This, combined with the principle of parallax, allows for millimeter-level positional accuracy. After confirming the file's location, the robotic arm follows a pre-set trajectory to remove the file from the shelf, grabbing and securing it before moving along a planned path to the outbound handover area. This area is equipped with an identity verification and file handover management system. After completing position calibration, the robot waits for the operator to verify their identity through biometric recognition, enabling the secure handover of the file entity. The entire process utilizes multi-sensor data fusion and intelligent control algorithms to ensure the efficiency and reliability of the removal operation.

[0100] The intelligent archive management method further includes, based on environmental control equipment, real-time monitoring of air quality parameters of the archive warehouse, and when the air quality parameters exceed a preset safety range, automatically adjusting the air quality parameters to the preset safety range through a fuzzy PID algorithm;

[0101] Based on surveillance cameras, a multimodal anomaly detection model is used to analyze the behavior of the operator in real time. If abnormal behavior is identified, an audible and visual alarm is issued and the entry and exit facial access control systems are locked;

[0102] Based on electronic archive scanning equipment, OCR and NLP technologies are used to automatically extract the archive text content in the archive warehouse, compare multiple archive text contents through data consistency verification algorithms, correct the contradictory archive text contents, and scan the archive text contents in real time through sensitive information recognition models. If sensitive information is found, the operation end is immediately notified.

[0103] Among them, air quality parameters specifically include temperature and humidity. The formula for temperature regulation using the fuzzy PID algorithm is as follows:

[0104] Where, Indicates the temperature control output value of the archive warehouse; Indicates the temperature setting value of the archive warehouse; Indicates the current measured temperature of the archive warehouse; represents the proportionality coefficient; Indicates the proportional coefficient correction amount; represents the integral coefficient; Indicates the integral coefficient correction amount; represents the differential coefficient; Indicates the correction amount of the differential coefficient.

[0105] First, an air quality monitoring network for archive warehouses can be established, relying on environmental control equipment, to collect parameters such as particulate matter concentration and harmful gas content in real time. When the monitoring data exceeds the preset safety threshold for archive preservation, an intelligent PID control strategy based on fuzzy logic is immediately activated. By linking equipment such as air purification systems and fresh air devices, ventilation volume and filtration efficiency are dynamically adjusted to stabilize air quality parameters within a standard range suitable for archive preservation, thus preventing environmental factors from damaging the archive entities.

[0106] Then, using surveillance cameras deployed in the warehouse, combined with a multimodal anomaly detection model that integrates visual features and behavioral time series, we can conduct real-time analysis of operators' movement trajectories, dwell time, and other behavioral characteristics. When the model identifies risky behaviors such as unauthorized intrusion or unusual dwelling, it immediately triggers the warehouse's audio and visual alarms, and links them to the biometric access control system, automatically locking entry and exit channels. This creates a closed-loop management mechanism for security incident detection, alarm, and resolution, ensuring the physical security of the archive warehouse.

[0107] Finally, electronic archival scanning equipment can be used to automatically extract and semantically parse text content from paper archives using OCR (Optical Character Recognition) and NLP (Natural Language Processing) technologies. A semantic association network of archival content can then be constructed based on the knowledge graph. Graph neural network algorithms can be used to verify the consistency of multi-source archival texts, and intelligently correct content that exhibits temporal inconsistencies or conflicting entity descriptions. Furthermore, a deep learning-based sensitive information recognition model can be deployed to scan electronic archival texts in real time. Once confidential information is detected, an alert notification is immediately sent to the operator, enabling secure management and quality optimization of archival data.

[0108] The step of comparing the plurality of archive text contents by using a data consistency check algorithm and correcting the conflicting archive text contents specifically includes:

[0109] Constructing an archival knowledge graph containing all the archival text contents based on a graph neural network, wherein each archival text content serves as a node of the archival knowledge graph, and the relationship between two archival text contents serves as an edge of the archival knowledge graph;

[0110] Node embedding vectors are learned through graph neural networks, and semantically related nodes are close in the vector space. The anomaly detection model based on the graph neural network calculates the node similarity of the two node embedding vectors. If the node similarity is lower than the preset similarity threshold, there is a potential contradiction between the two nodes, and a rule-based secondary verification of the two nodes is required. When a contradiction is detected again, a revised version of the archive content is selected based on the node credibility and a corresponding archive content correction suggestion is generated;

[0111] Among them, the loss function L of the anomaly detection model of the graph neural network is:

[0112] Where L represents the loss function of the anomaly detection model of the graph neural network; V represents the node set of the archive knowledge graph; represents a node in the node set V; E represents the edge set of the archive knowledge graph; represents an edge in the edge set E; Represents nodes respectively The node embedding vector of Represents two node embedding vectors Node similarity; represents cosine similarity; Represents the node embedding vector The transpose of Represents the node embedding vector The model; Represents the node embedding vector The model; represents the sigmoid activation function; W represents the learnable weight matrix; b represents the bias term; Represents the node embedding vector and splicing operation; Represents an edge The true label of , if the content of nodes u and v is the same, then , if the contents of nodes u and v are contradictory, then ; represents the L2 regularization term; represents the regularization strength parameter, ; Represents the set of all learnable parameters in a graph neural network;

[0113] The transfer function of each layer of the graph neural network is:

[0114] Where, Represents the node embedding vector of node u in the l+1 layer; Represents the node embedding vector of node u in layer l; represents the learnable weight matrix of layer l; represents the set of neighbor nodes of node u, ; represents the symmetric normalization constant, , represents the degree of node u, that is, the number of neighbor nodes, Represents the degree of node v, that is, the number of neighbor nodes.

[0115] First, based on graph neural network technology, all archival text content can be constructed into a structured archival knowledge graph. This graph uses individual archival texts as basic nodes and semantic connections between archives, such as event references, chronological order, and person associations, as connecting edges, forming a multidimensional semantic network. Through this modeling approach, archival content is transformed from discrete text data into knowledge entities with a topological structure, providing a semantic association foundation for subsequent consistency analysis.

[0116] The multi-layer transfer mechanism of graph neural networks can then be used to learn semantic vectorized representations of nodes in the archival knowledge graph. Each network layer updates the embedding vector of the current node by aggregating feature information from neighboring nodes. This process begins by collecting features from all neighboring nodes, transforming these features using a learnable weight matrix, and then balancing the weights of nodes of varying degrees using a symmetric normalization constant. Finally, an activation function introduces nonlinear transformations to generate higher-level semantic representations. This layer-by-layer transfer mechanism enables node embedding vectors to incorporate semantic information from multi-hop neighbors, enabling deep semantic feature extraction of archival content.

[0117] Furthermore, the semantic consistency of archive nodes can be assessed using a dual similarity calculation model. First, cosine similarity is used to measure the directional consistency of node embedding vectors. A similarity score is then generated through a nonlinear transformation of the concatenated vectors and learnable parameters. The two are combined to form the final node similarity metric. A semantic similarity threshold can also be preset. When the similarity between two nodes falls below this threshold, a potential contradiction is identified. For initially detected contradictory pairs, a rule-based secondary verification mechanism is activated. This mechanism uses domain rules such as temporal logic verification and entity relationship constraints to further confirm the authenticity of the contradiction and avoid misjudgments.

[0118] For discrepancies in archive content identified through secondary verification, a revised version can be selected based on a node credibility assessment model. This credibility assessment comprehensively considers metadata such as the archive's creation date, source authority, and modification history, prioritizing highly credible node content as the basis for revisions. A structured revision suggestion is also generated, including the location of the discrepancy, a comparison of the pre- and post-correction content, and an explanation of the revision rationale for the revision, for review and reference by the operator. This mechanism enables intelligent processing throughout the entire process, from discrepancy detection to content revision.

[0119] This data consistency verification technology, through the deep integration of knowledge graphs and graph neural networks, breaks through the semantic understanding limitations of traditional text comparison methods and enables semantic-level contradiction detection in archival content. Multi-layer graph feature learning can capture the implicit deep connections between archives, and the dynamic similarity calculation model adapts to the semantic expression characteristics of different types of archives. The credibility-driven correction mechanism ensures data accuracy while retaining the decision space for manual review. In its application at a provincial archive, this data consistency verification technology has increased the accuracy of archival content contradiction detection to 92%, significantly improving archival data quality.

[0120] Furthermore, to improve the accuracy of contradiction detection, a loss function consisting of a reconstruction error term and a regularization term can be used to optimize the training of the graph neural network anomaly detection model. The reconstruction error term measures the difference between the model's predicted similarity and the true label. By minimizing this error, semantically related archive nodes are moved closer together in the vector space, while contradictory nodes are moved further away. The L2 regularization term constrains the complexity of the model parameters, preventing overfitting and improving the model's generalization ability. The regularization strength parameter is dynamically adjusted based on the size and complexity of the archive data to ensure the model's stability in different application scenarios.

[0121] The model is trained using a loss function that includes a reconstruction error and a regularization term. By minimizing the difference between predicted similarity and true labels, it enhances the clustering of semantically similar nodes and suppresses the erroneous association of conflicting nodes. Furthermore, a symmetric normalization mechanism balances the influence of nodes of varying degrees, preventing highly connected nodes from dominating embedding learning and ensuring a balanced graph representation. This mechanism enables semantic-level contradiction detection in archival texts, improving contradiction identification accuracy by 30% compared to traditional rule-based verification, providing intelligent support for high-quality archival data management.

[0122] like Figure 3 FIG. 1 is a system block diagram of an intelligent archive management system provided by an embodiment of the present invention, wherein the management system includes:

[0123] Intelligent RFID electric compact cabinet, used for storing files, including compact shelves, drive motors, RFID readers and positioning sensors;

[0124] The retrieval robot is used to perform file shelving and de-shelving tasks, including a mobile chassis, a robotic arm, a vision system, an RFID reader, and a battery management system;

[0125] Facial access control equipment, used for personnel identity verification, including cameras, facial recognition devices and access control controllers;

[0126] Environmental control equipment, used to adjust the air quality parameters of the archive warehouse, including temperature and humidity sensors, air conditioners, dehumidifiers and ventilation systems;

[0127] Monitoring equipment for security monitoring of the archive warehouse, including cameras, video analysis servers, and sound and light alarms;

[0128] Electronic archival scanning equipment, used for digital scanning of archival text content, including a high-speed scanner, image processing module, and OCR recognition module;

[0129] Intelligent interactive terminal, used for interaction between the operating terminal and the intelligent archive management system, including touch screen, input device and display device;

[0130] Database server, used to store archival information, equipment status and operation records.

[0131] Figure 3 The apparatus of the embodiment shown can be used to perform Figure 1 The implementation principles and technical effects of the steps in the method embodiment shown are similar and will not be repeated here.

[0132] The schematic diagram of the smart RFID electric compact cabinet is as follows Figure 4 shown.

[0133] An electronic device includes a memory and a processor, wherein the memory stores a computer program. When the processor runs the computer program stored in the memory, the processor executes the steps of any one of the above-described intelligent archive management methods.

[0134] like Figure 5 FIG. 1 is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. The electronic device 50 includes: a processor 51, a memory 52 and a computer program;

[0135] The memory 52 is used to store the computer program, which may also be a flash memory. The computer program is, for example, an application program or a functional module for implementing the above method.

[0136] The processor 51 is configured to execute the computer program stored in the memory to implement the various steps performed by the device in the above method. For details, please refer to the relevant description in the above method embodiment.

[0137] Optionally, the memory 52 may be independent or integrated with the processor 51 .

[0138] When the memory 52 is a device independent of the processor 51, the device may further include:

[0139] The bus 53 is used to connect the memory 52 and the processor 51 .

[0140] A readable storage medium stores a computer program, which, when executed by a processor, is used to implement the steps of any one of the above-described intelligent archive management methods.

[0141] The readable storage medium may be a computer storage medium or a communication medium. Communication media include any medium that facilitates the transfer of computer programs from one location to another. Computer storage media may be any available medium that can be accessed by a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium may also be an integral part of the processor. The processor and the readable storage medium may be located in an application-specific integrated circuit (ASIC). In addition, the ASIC may be located in a user device. Of course, the processor and the readable storage medium may also exist as discrete components in a communication device. The readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0142] The present invention also provides a program product, which includes execution instructions stored in a readable storage medium. At least one processor of a device can read the execution instructions from the readable storage medium, and at least one processor executes the execution instructions so that the device implements the methods provided in the various embodiments described above.

[0143] In the embodiments of the above-mentioned devices, it should be understood that the processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASICs), etc. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly executed by a hardware processor or by a combination of hardware and software modules within the processor.

[0144] Through the introduction of the above embodiments, the present invention can establish a three-dimensional archive location information map of the archive warehouse based on BIM technology through the intelligent archive management method and system, initialize the archive location database and configure the equipment communication protocol; after the operation end verifies the identity through the face access control of the entry, the archive to be put on the shelf is placed in the entry temporary storage area; according to the RFID archive tag of the archive to be put on the shelf, the matching target archive shelving location is retrieved in the archive location database and an archive shelving instruction is generated; the archive shelving instruction is received, and the retrieval and return robot is dispatched to perform the archive shelving task according to the target archive shelving location; after the operation end inputs the retrieval query instruction through the intelligent interactive terminal, the archive to be taken off the shelf is determined, and according to the RFID archive tag of the archive to be taken off the shelf, the matching target archive off the shelf location is retrieved in the archive location database and an archive off the shelf instruction is generated; the archive off the shelf instruction is received, and the retrieval and return robot is dispatched to perform the archive off the shelf task according to the target archive off the shelf location; after the archive off the shelf task is completed, the operation end verifies the identity through the face access control of the exit A distributed inventory strategy is used in the archive warehouse to scan and identify archives, generate archive storage location records, update the archive location database, and generate archive inventory reports. Based on environmental control equipment, the air quality parameters of the archive warehouse are monitored in real time. When the air quality parameters exceed the preset safety range, the fuzzy PID algorithm is used to automatically adjust the air quality parameters to the preset safety range; based on surveillance cameras, a multimodal anomaly detection model is used to analyze the behavior of the operator in real time. If abnormal behavior is identified, an audible and visual alarm is issued and the entry and exit facial access control systems are locked; based on electronic archive scanning equipment, OCR and NLP technologies are used to automatically extract the archive text content of the archive warehouse, compare multiple archive text contents through a data consistency verification algorithm, correct conflicting archive text contents, and scan the archive text content in real time through a sensitive information recognition model. If sensitive information is found, the operator is immediately notified, thereby solving the problems of multiple manual operations, low efficiency, and incomplete information management in the existing technology, significantly improving the efficiency of archive access, and ensuring the accuracy, security and intelligence of archive management.

[0145] The present invention uses an intelligent archive management system composed of intelligent RFID electric compact cabinets, archive retrieval and return robots, etc., to realize the automation of archive management, improve management efficiency, reduce the risk of manual operation, and significantly improve the efficiency of archive access compared to existing technologies. The present invention can integrate UWB positioning technology and visual recognition technology to accurately determine the coordinates and status of storage slots, effectively improve the accuracy and security of archive access, and reduce the risk of archive loss or misplacement. The present invention can realize the electronic management and real-time updating of archive information, comprehensively record the archive life cycle, improve the intelligence level of archive management, and meet the needs of high-efficiency and intelligent archive management.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent file management method, characterized in that: The management method includes: Based on BIM technology, a three-dimensional archive location information map of the archive warehouse was established, the archive location database was initialized, and the equipment communication protocol was configured; After the operator verifies his / her identity through the facial recognition access control system, he / she places the files to be put on the shelves in the temporary storage area; based on the RFID file tags of the files to be put on the shelves, he / she searches the file location database for matching target file shelving locations and generates file shelving instructions; he / she receives the file shelving instructions and dispatches the file retrieval and return robot to perform the file shelving task according to the target file shelving location; The operating terminal inputs a file retrieval query instruction through the intelligent interactive terminal to determine the file to be removed from the shelf, and searches the file location database for a matching target file removal location based on the RFID file tag of the file to be removed from the shelf and generates a file removal instruction; receives the file removal instruction, and dispatches the file retrieval and return robot to perform the file removal task based on the target file removal location; after the file removal task is completed, the operating terminal verifies its identity through the outbound facial access control and takes away the file to be removed from the outbound handover area; According to the file inventory instruction, the file retrieval robot is dispatched to scan and identify files in the file warehouse using a distributed inventory strategy, generate a file storage location record and update the file location database, and generate a file inventory report; The system further includes monitoring the air quality parameters of the archive warehouse in real time based on environmental control equipment, and automatically adjusting the air quality parameters to the preset safety range through a fuzzy PID algorithm when the air quality parameters exceed a preset safety range; Based on surveillance cameras, a multimodal anomaly detection model is used to analyze the behavior of the operator in real time. If abnormal behavior is identified, an audible and visual alarm is issued and the entry and exit facial access control systems are locked; Based on electronic file scanning equipment, OCR and NLP technologies are used to automatically extract the text content of the archives in the archive warehouse, compare multiple archive texts through a data consistency verification algorithm, correct the inconsistent archive texts, and scan the archive texts in real time through a sensitive information recognition model. If sensitive information is found, the operation terminal is immediately notified; The step of comparing the plurality of archive text contents by using a data consistency check algorithm and correcting the conflicting archive text contents specifically includes: Constructing an archival knowledge graph containing all the archival text contents based on a graph neural network, wherein each archival text content serves as a node of the archival knowledge graph, and the relationship between two archival text contents serves as an edge of the archival knowledge graph; Node embedding vectors are learned through graph neural networks, and semantically related nodes are close in the vector space. The anomaly detection model based on the graph neural network calculates the node similarity of the two node embedding vectors. If the node similarity is lower than the preset similarity threshold, there is a potential contradiction between the two nodes, and a rule-based secondary verification of the two nodes is required. When a contradiction is detected again, a revised version of the archive content is selected based on the node credibility and a corresponding archive content correction suggestion is generated; Among them, the loss function L of the anomaly detection model of the graph neural network is: L=∑ (u,v)∈E (Sim(h u ,h v )-m u,v ) 2 +γ||θ|| 2 Where L represents the loss function of the anomaly detection model of the graph neural network; V represents the node set of the archive knowledge graph; u, v represent the nodes in the node set V; E represents the edge set of the archive knowledge graph; (u, v) represents the edge in the edge set E; h u ,h v Represents the node embedding vector of node u,v respectively; Sim(h u ,h v ) represents two node embedding vectors h u ,h v Node similarity; represents cosine similarity; Represents the node embedding vector h u The transpose of ||h u || represents the node embedding vector h u The module of ||h v || represents the node embedding vector h v The modulus of ; σ represents the sigmoid activation function; W represents the learnable weight matrix; b represents the bias term; Represents the node embedding vector h u and h v splicing operation; m u,v represents the true label of edge (u,v). If the content of nodes u and v is the same, then g u,v =1, if the contents of nodes u and v are contradictory, then g u,v =0;γ||θ|| 2 represents the L2 regularization term; γ represents the regularization strength parameter, 0.0001≤γ≤0.01; θ represents the set of all learnable parameters in the graph neural network; The transfer function of each layer of the graph neural network is: Where, Represents the node embedding vector of node u in the l+1 layer; represents the node embedding vector of node u in layer l; W (l) represents the learnable weight matrix of layer l; N(u) represents the set of neighbor nodes of node u, N(u) = {v∈V|(u,v)∈E or (v,u)∈E}; p u,v represents the symmetric normalization constant, |N(u)| represents the degree of node u, that is, the number of neighboring nodes, and |N(v)| represents the degree of node v, that is, the number of neighboring nodes.

2. The intelligent file management method according to claim 1, characterized in that: Configuring a unique file identifier for each file, and binding the unique file identifier to the file storage location record via the RFID tag; Among them, the unique file identifier is divided into a person file identifier and an event file identifier. The person file identifier includes the person's name, gender and ID number, and the event file identifier includes the location of the event, the timestamp of the event and the event category code.

3. The intelligent file management method according to claim 1, characterized in that: The step of scheduling the file retrieval and return robot to perform the file shelving task according to the target file shelving location specifically includes: Constructing a dynamic storage optimization model, and dispatching the file retrieval and return robot to perform the file shelving task based on the file importance score of the file to be shelved; The file retrieval and return robot plans the file shelving task path by improving the A* algorithm, moves to the storage temporary storage area based on the file shelving task path and grabs the file to be shelved, then moves to the target file shelving position, combines UWB positioning technology and visual recognition technology to determine the file shelving grid, and places the file to be shelved in the file shelving grid.

4. The intelligent file management method according to claim 3, characterized in that: The heuristic function H(n) of the improved A* algorithm is as follows: Where H(n) represents the heuristic function of the improved A* algorithm, that is, the estimated cost of the retrieval robot from the node n corresponding to the current position to the node g corresponding to the target file shelf position; (x n ,y n ,z n ) represents the coordinates of node n in the three-dimensional archive location information map; (x g ,y g ,z g ) represents the coordinates of node g in the three-dimensional archive location information map; represents the Euclidean distance from node n to node g of the retrieval and return robot, i.e., the straight-line distance; α represents the path complexity weight coefficient, 0.5<α<2.0; α0 represents the basic weight coefficient, α0=1.0; β represents the adjustment coefficient, 0.5<β<1.5; ObstacleDensity represents the obstacle density of the archive warehouse; MaxDensity represents the maximum obstacle density of the archive warehouse; PathDifficulty(n) represents the comprehensive difficulty evaluation value of the path where node n is located; ω i represents the weight coefficient of the i-th path impact factor of the path where node n is located; f i (n) represents the i-th path impact factor of the path where node n is located; f1(n) represents the first path impact factor of the path where node n is located, that is, the channel width impact factor; w(n) represents the real-time measurement width of the channel where node n is located; w min Indicates the minimum aisle width of the archive warehouse; w max represents the maximum channel width of the archive warehouse; f2(n) represents the second path impact factor of the path where node n is located, that is, the obstacle impact factor; d i represents the Euclidean distance from node n to the ith obstacle; m represents the total number of obstacles within a radius of 5 meters centered on node n; λ represents the attenuation coefficient, λ = 0.5; f3(n) represents the third path influence factor of the path where node n is located, that is, the illumination influence factor; I thershold represents the minimum working light intensity of the visual system of the retrieval and return robot; I(n) represents the light intensity at node n; ∈ represents a constant, ∈ = 1.0; f4(n) represents the fourth path influence factor of the path where node n is located, that is, the motion state factor of the compact shelving; μ represents the speed influence coefficient; v(n) represents the motion speed of the compact shelving in the channel where node n is located.

5. The intelligent file management method according to claim 3, characterized in that: The center coordinates (x final ,y final ,z final )for: In the formula, (x final ,y final ,z final ) represents the center coordinates of the archive shelving grid in the three-dimensional archive location information map; q uwb Represents the confidence weight of UWB positioning; q vis Represents the confidence weight of visual recognition; (x uwb ,y uwb ,z uwb ) represents the estimated coordinate value of the archive shelving grid provided by the UWB positioning technology in the three-dimensional archive location information map, which is obtained by triangulation using the UWB positioning base station deployed in the archive warehouse; (x vis ,y vis ,z vis ) represents the estimated coordinate value of the archive shelving grid extracted by visual recognition technology in the three-dimensional archive storage location information map, which is obtained by recognizing the grid visual identifier through the camera on the retrieval and return robot and calculating it through perspective visual transformation; ξ uwb represents the standard deviation of UWB positioning error, which is determined by the UWB positioning base station deployment density and the environmental multipath effect; τ represents a constant, τ = 10 -6 ConfidenceScore represents the confidence level of the grid recognition output by the visual recognition technology; MaxScore represents the maximum possible confidence level of the visual recognition technology; ξ vis represents the standard deviation of the visual positioning error; r represents the distance from the camera on the retrieval and return robot to the file shelving opening; δ represents the camera pixel error; and c represents the focal length of the camera.

6. The intelligent file management method according to claim 5, characterized in that: Calculate the location verification threshold of the file shelf slot ValidationThreshold = k*(q uwb +q vis ), where k represents the position safety factor; When ||(x final ,y final ,z final )-(x ref ,y ref ,z ref )||>ValidationThreshold, it is necessary to determine the center coordinates of the file shelf opening again, where (x ref ,y ref ,z ref ) represents the reference coordinates of the archive shelving grid stored in the grid database.

7. The intelligent file management method according to claim 1, characterized in that: The receiving of the file removal instruction and scheduling the file retrieval and return robot to perform the file removal task according to the target file removal location specifically includes: After receiving the file removal instruction, the file retrieval robot with sufficient power and closest to the target file removal location is preferentially dispatched to perform the file removal task; The file retrieval and return robot plans the file unloading task path through the improved A* algorithm, moves to the target file unloading position based on the file unloading task path, adjusts the gripping force of the robotic arm through the force control algorithm, and determines the target file unloading position in combination with the binocular vision mechanism. The robotic arm takes down the file to be unloaded and moves it to the outbound handover area.

8. An intelligent file management system, applied to the intelligent file management method according to any one of claims 1 to 7, characterized in that: The management system includes: Intelligent RFID electric compact cabinet, used for storing files, including compact shelves, drive motors, RFID readers and positioning sensors; The retrieval robot is used to perform file shelving and de-shelving tasks, including a mobile chassis, a robotic arm, a vision system, an RFID reader, and a battery management system; Facial access control equipment, used for personnel identity verification, including cameras, facial recognition devices and access control controllers; Environmental control equipment, used to adjust the air quality parameters of the archive warehouse, including temperature and humidity sensors, air conditioners, dehumidifiers and ventilation systems; Monitoring equipment for security monitoring of the archive warehouse, including cameras, video analysis servers, and sound and light alarms; Electronic archival scanning equipment, used for digital scanning of archival text content, including a high-speed scanner, image processing module, and OCR recognition module; Intelligent interactive terminal, used for interaction between the operating terminal and the intelligent archive management system, including touch screen, input device and display device; Database server, used to store archival information, equipment status and operation records.

Images