Intelligent identification method and system for warehousing and unwarehousing of figure ammunition equipment

Abstract

The invention, which relates to the technical field of intelligent warehouse management, discloses an intelligent identification method and system for warehouse-in and warehouse-out of weather modification ammunition equipment. Warehouse-out and warehouse-in operations are performed based on a warehouse-out demand table, warehouse-out information, a warehouse-out state and warehouse-in application information by performing inventory registration on each piece of weather modification ammunition equipment in an equipment warehouse; the method comprises the following steps: performing visual image tracking on a plurality of warehousing-in and warehousing-out weather modification ammunition equipment, obtaining a plurality of frames of equipment orientation pictures based on an image tracking result, obtaining warehousing-in and warehousing-out trace information according to all frames of equipment orientation pictures, analyzing and extracting to obtain a warehousing-out feature set and a warehousing-in feature set, and comparing the warehousing-out feature set and the warehousing-in feature set with a pre-configured warehousing-in and warehousing-out template; according to the method, abnormal warehouse-in and warehouse-out behaviors are judged based on a comparison result, information extraction is carried out on the warehouse-in and warehouse-out behaviors, a topological information tree is established to carry out linkage information pushing, and safe, efficient and accurate ammunition equipment warehouse-in and warehouse-out management is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent warehouse management, in particular to a human shadow ammunition equipment warehouse-in and warehouse-out intelligent identification method and system. BACKGROUND

[0002] In the traditional warehouse management method, especially the warehouse-in and warehouse-out management of special materials (such as human shadow ammunition equipment), most of them rely on manual recording, barcode scanning or simple visual identification technology for inventory registration and tracking, which is easy to miss or misrecord, and difficult to synchronize with the real warehouse-in and warehouse-out operation in time, resulting in inaccurate inventory information. The existing visual identification is often limited to local node shooting, lacks continuous tracking capability, and it is difficult to obtain complete warehouse-in and warehouse-out trajectory information. Moreover, it lacks multi-dimensional feature extraction and comparison of warehouse-in and warehouse-out behaviors, resulting in that various information generated in the warehouse-in and warehouse-out process cannot be effectively associated and integrated, leading to difficulty in tracing and insufficient linkage response capability, thereby affecting the efficiency of equipment warehouse-in and warehouse-out. Therefore, there is an urgent need for a warehouse-in and warehouse-out management method that can realize full-process visual tracking, feature extraction, abnormal behavior automatic identification and information linkage push, so as to improve the safety, accuracy and efficiency of human shadow ammunition equipment management. SUMMARY

[0003] The purpose of the present application is to provide a human shadow ammunition equipment warehouse-in and warehouse-out intelligent identification method and system to solve the problems in the background art.

[0004] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a human shadow ammunition equipment warehouse-in and warehouse-out intelligent identification method, comprising the following steps: Step S1: inventory registration is performed for each human shadow ammunition equipment in the equipment warehouse, warehouse-out operation is performed based on the warehouse-out demand table and inventory information, and warehouse-in operation is performed based on the inventory state and warehouse-in application information; Step S2: visual image tracking is performed on the several human shadow ammunition equipment in and out of the warehouse, several frames of equipment directional pictures are obtained based on the image tracking results, and warehouse-in and warehouse-out trace information is obtained according to the equipment directional pictures of all frames; Step S3: the warehouse-in and warehouse-out trace information is analyzed, the warehouse-out feature set and the warehouse-in feature set are extracted, and compared with the pre-configured warehouse-in and warehouse-out template, abnormal warehouse-in and warehouse-out behaviors are identified based on the comparison results, information extraction is performed on the warehouse-in and warehouse-out behaviors, a topological information tree is established, and linkage information push is performed.

[0005] In a preferred embodiment, the process of inventory registration for each human shadow ammunition equipment in the equipment warehouse, warehouse-out operation based on the warehouse-out demand table and inventory information, and warehouse-in operation based on the inventory state and warehouse-in application information comprises: The system collects equipment identification information for each type of artificial rainmaking ammunition and equipment, creates its own registration text, stores all registration texts in the database, reviews the registration texts, completes the inventory registration of artificial rainmaking ammunition and equipment based on the review results, and handles the rejection of registration texts. The outbound demand table records the required manned ammunition and equipment to be issued. Data monitoring is performed on all registered texts stored in the database and approved to obtain the inbound and outbound records of manned ammunition and equipment in each equipment area. Inventory information is obtained based on the inbound and outbound records. According to the outbound demand list, select the required artificial rainmaking ammunition and equipment from the equipment warehouse and issue them out, and supervise the entire outbound process. According to the inbound demand list, transport the artificial rainmaking ammunition and equipment to the equipment warehouse and supervise the entire inbound process.

[0006] In a preferred embodiment, the process of visually tracking several pieces of artificial intelligence ammunition equipment entering and leaving the warehouse includes: Several key storage location nodes are set on the outbound and inbound routes, covering the storage door area, shelf aisles, and loading and unloading points. The storage door area is used as the visual starting point and the loading and unloading point is used as the visual ending point to perform visual image tracking of the entire outbound operation process. The positions of the visual starting point and visual ending point are swapped to perform visual image tracking of the entire inbound operation process. Set up several outbound tracking nodes and deploy several video recording devices at the visual starting point, visual ending point and outbound tracking nodes. The video recording devices capture several frames of video clips of equipment leaving the warehouse. By connecting different video clips, an outbound scene animation is obtained as the video tracking result of the outbound operation. Several inbound tracking nodes are set up, and the corresponding robotic arm operation points are set up for deploying the robotic arm. Image recording devices are deployed at the inbound tracking nodes, and imaging capture areas are set up to obtain inbound image frames of the manned ammunition equipment at each position. When the manned ammunition equipment at any inbound tracking node is not in the imaging capture area, the robotic arm at the current inbound tracking node adjusts the position of the manned ammunition equipment to the imaging capture area. The artificial ammunition and equipment are loaded from the loading and unloading point and stored in the equipment depot. All the images of the entire process of the storage operation are integrated as the image tracking results of the storage operation.

[0007] In a preferred embodiment, the process of obtaining several frames of equipment orientation images based on image tracking results includes: Set up a first processing window and a second processing window; The first processing window is used to perform image preprocessing on several image segments and imported image frames, including image denoising, image enhancement and layered distortion correction; a standard noise image is obtained through image denoising, the standard noise image is converted into a standard fitted image through image enhancement, and layered distortion correction is performed on the standard fitted image to obtain the equipment frame interface map, which is covered with a complete image of the artificial munitions and equipment as well as the surrounding environment. The second processing window is used to reconstruct the equipment frame interface image, divide the equipment frame interface image into several reconstruction areas, and identify the image features under each reconstruction area. The image features include directional element features and non-directional element features. When a certain reconstructed region contains only directional element features, the reconstructed region is retained as the object to be stitched. When it contains only non-directional element features, the reconstructed region is discarded. When both directional and non-directional element features exist, the non-directional element features are stripped from the reconstructed region, and the reconstructed region containing only directional element features is taken as the object to be stitched. All the objects to be stitched are stitched together according to the time order and position order to obtain several frames of equipment directional images.

[0008] In a preferred embodiment, the process of obtaining entry / exit trace information based on the equipment orientation images of all frames includes: In each frame of the equipment orientation screen, select several trace calibration points, connect the trace calibration points in the same equipment position in different frames of the equipment orientation screen to obtain the entry and exit trace path, and divide the entry and exit trace path into several information nodes based on the operation timestamp. Import the operation-related information of the artificial ammunition equipment at the current time and location into each information node. Integrate the operation-related information of all information nodes on the entry and exit trace path and the evidence snapshot link into the entry and exit trace information. Encapsulate the entry and exit trace information into a data packet and map the address of the data packet to a pre-deployed virtual machine.

[0009] In a preferred embodiment, the image enhancement and layered distortion correction process includes: Several regions to be enhanced are defined in a standard noisy image. Pixel contrast, illumination intensity, and color saturation are used as enhancement type items. Pixel blocks that do not meet the requirements of any enhancement type item are obtained and marked as enhancement points under the regions to be enhanced. Group all points to be enhanced under the same enhancement type into a type point set, set several gradient enhancement intervals, integrate the points to be enhanced under the same type point set into a gradient point cluster, perform pixel adjustment on several gradient point clusters, and convert the standard noisy image into a standard fitted image. The standard fitted image is divided into different focus plane layers based on the depth of field. The standard fitted images of the same focus plane layer are integrated into a layered image set. A sliding correction window is set to traverse the window of each layered image set, and the standard fitted image is divided into several sub-regions to be corrected. In each sub-region to be corrected, a number of distorted points are marked and the distortion type of the distorted points is labeled. Distorted points of the same distortion type are grouped into a distortion set. Based on the number of distortion sets, a corresponding number of correction parameters are set. Each correction parameter is used to correct the distorted points under the corresponding distortion set, thereby converting the distorted points into normal points. After completing the distortion correction of all layered images, all standard fitted images are converted into equipment frame interface images.

[0010] In a preferred embodiment, the process of analyzing inbound and outbound trace information, extracting outbound feature sets and inbound feature sets, comparing them with pre-configured inbound and outbound templates, and identifying abnormal inbound and outbound behaviors based on the comparison results includes: Several types of feature extraction rules are set to extract features from the entry and exit trace information, resulting in different types of entry and exit related features, including operation context features, spatiotemporal trajectory features, equipment identity features, and behavior sequence features. Based on the entry and exit trace information, determine the operation type of the current artificial ammunition equipment. When the operation type is entry, extract all entry and exit related features related to the current entry to generate an entry feature set. When the operation type is exit, extract all entry and exit related features related to the current exit to generate an exit feature set. An inbound / outbound template is pre-configured, which includes several text recognition words. A rule engine is deployed to access several feature extraction rules, inbound / outbound related features, and the inbound / outbound template. All operational behaviors of each manned ammunition and equipment during the inbound / outbound process are compared sequentially. If the inbound / outbound feature of an operational behavior fails to match the text recognition words of the inbound / outbound template, the operational behavior is filtered as an abnormal inbound / outbound behavior until all abnormal inbound / outbound behaviors are filtered out.

[0011] In a preferred embodiment, the process of extracting information from inbound and outbound activities includes: The default information rule base is used to store rule forms and identifier forms; Based on the information rule base, all inbound and outbound behaviors are grouped by type and labeled by event. The type grouping includes primary grouping and secondary grouping, and the rule form includes initial screening form and re-screening form. The initial screening form is used to group inbound and outbound behaviors once and filter out all abnormal inbound and outbound behaviors. The secondary screening form groups the abnormal inbound and outbound behaviors a second time and obtains the abnormal details type for each abnormal inbound and outbound behavior. Normal inbound and outbound behaviors are not processed. The abnormal details types include quantity abnormality, time sequence abnormality, and frequency abnormality. The identification form is used to annotate abnormal inbound and outbound behaviors for different abnormal detail types. Each abnormal inbound and outbound behavior is labeled as a standardized abnormal structure event, and the event information is labeled, including event ID, abnormal type, timestamp, core entity, and event details.

[0012] In a preferred embodiment, the process of establishing a topology information tree for linked information push includes: All event nodes and event edges are connected based on the relationship between the mapping elements to build an information subtree. A root node is set, and all information subtrees are connected to the root node to obtain the topology information tree. When a new standardized abnormal structure event is added, the information subtree corresponding to the current standardized abnormal structure event is built and mounted on the topology information tree. When performing an information retrieval of a certain artificial ammunition equipment entering or leaving the warehouse, the corresponding event ID and search terms are entered. The corresponding information subtree is retrieved from the topological information tree based on the event ID, and the corresponding event node on the information subtree is retrieved based on the search terms. The event information of the directly retrieved event nodes will be used as the direct demand information; Traverse all other event nodes on the event edge of the information subtree corresponding to the retrieved event node, take the event information of other event nodes as linkage information, divide the linkage information into depth levels, and fold the linkage information into several depth levels of data linkage layers. Based on the depth hierarchy, the folded linkage information within the data linkage layer is unfolded sequentially from low to high. The object performing the information retrieval decides whether to select the linkage information within the data linkage layer. Based on the directly retrieved information and the linkage information, decision-making is made regarding abnormal inbound and outbound behaviors.

[0013] This invention also provides an intelligent identification system for the entry and exit of manned ammunition and equipment, the system comprising: The inbound / outbound module is used to register the inventory of each manned ammunition and equipment in the equipment warehouse, perform outbound operations based on the outbound demand table and inventory information, and perform inbound operations based on inventory status and inbound application information. The visual tracking module is used to perform visual image tracking on several pieces of artificial ammunition equipment entering and leaving the warehouse. Based on the image tracking results, several frames of equipment orientation images are obtained, and the entry and exit trace information is obtained based on all the equipment orientation images. The intelligent identification and information push module is used to analyze the entry and exit trace information, extract the exit feature set and the entry feature set, compare them with the pre-configured entry and exit templates, identify abnormal entry and exit behaviors based on the comparison results, extract information from the entry and exit behaviors, establish a topology information tree, and push related information.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention establishes inventory registration for each piece of equipment and performs full-process visual image tracking, which can accurately capture every frame of the equipment during the entry and exit process, forming a complete equipment orientation image sequence. This enables accurate restoration and digital preservation of the physical operation process. By analyzing the entry and exit trace information, the invention extracts the exit feature set and the entry feature set, and intelligently compares them with the preset entry and exit templates. This can automatically identify abnormal behaviors such as abnormal quantity, abnormal path, and inconsistent identity, thereby improving the level of safety management to a certain extent.

[0015] 2. This invention establishes a topological information tree to hierarchically organize and push abnormal events and related information. When users search for information, they can obtain information that meets their direct needs or expand related information, thereby providing comprehensive and structured data support for the analysis, tracing, and decision-making of abnormal behavior, effectively improving decision-making efficiency, and indirectly improving the management efficiency of ammunition and equipment entering and leaving the warehouse. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of the method of the present invention.

[0018] Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1, please refer to Figure 1As shown in the figure, the intelligent identification method for the entry and exit of artificial ammunition equipment described in this embodiment includes the following steps: Step S1: Register the inventory of each manned ammunition equipment in the equipment warehouse, perform the outbound operation based on the outbound demand table and inventory information, and perform the inbound operation based on the inventory status and inbound application information. Step S2: Perform visual image tracking on several pieces of artificial ammunition equipment entering and leaving the warehouse, obtain several frames of equipment orientation images based on the image tracking results, and obtain entry and exit trace information based on all frames of equipment orientation images. Step S3: Analyze the entry and exit trace information, extract the exit feature set and the entry feature set, compare them with the pre-configured entry and exit templates, identify abnormal entry and exit behaviors based on the comparison results, extract information from the entry and exit behaviors, establish a topology information tree for linkage information push.

[0021] It should be further explained that, in the specific implementation process, the process of registering the inventory of each artificial weather modification ammunition and equipment in the equipment depot, performing outbound operations based on the outbound demand form and inventory information, and performing inbound operations based on inventory status and inbound application information includes: The equipment identification information of all artificial reconnaissance ammunition equipment in the equipment warehouse is statistically analyzed. The equipment identification information specifically includes the storage location of the artificial reconnaissance ammunition equipment, the time of entry into the warehouse, different equipment types, and the number of equipment corresponding to each equipment type. Based on the equipment identity information, a registration text is created for each piece of artificial reconnaissance ammunition equipment. The registration text is used to record the entry and exit of artificial reconnaissance ammunition equipment. All registration texts are stored in the database deployed in the equipment warehouse. The database reviews the registration text. If the review is successful, the inventory registration of the current Munitions and Equipment is completed, and the corresponding equipment identification information is written into the registration text. If the review is unsuccessful, the current registration text is rejected. The review is based on the IP address associated with the registered text. If the IP address is not in the preset address list, the review will fail; otherwise, the review will pass. The outbound demand form is used to record all the artificial weathering ammunition and equipment that need to be outbound. Set an update frequency, and based on the update frequency, perform real-time data monitoring on all registered texts stored in the database that have been reviewed, thereby obtaining the entry and exit records of artificial ammunition equipment in each equipment area of ​​the current equipment warehouse. Inventory information is obtained based on all inbound and outbound records. The inventory information is used to represent the storage status of artificial ammunition and equipment in all equipment areas of the equipment depot at a certain time, including both equipment outbound and equipment in the depot. According to the outbound demand list, select the required artificial weathering ammunition and equipment from the equipment warehouse, carry out the outbound operation, and supervise the entire outbound process. According to the inbound demand list, transport the artificial weathering ammunition and equipment from outside to the equipment warehouse, carry out the inbound operation, and supervise the entire inbound process.

[0022] The inventory status is described as follows: Set an inventory alarm threshold. Based on the inventory information, obtain the number of artificial ammunition and equipment stored in the current equipment warehouse. When the number is less than or equal to the inventory alarm threshold, update the inventory status to "inventory depleted". Otherwise, mark the inventory status as "inventory normal".

[0023] It should be further explained that, in the specific implementation process, the visual image tracking of several pieces of artificial intelligence ammunition and equipment entering and leaving the warehouse includes: Several key storage location nodes are set on the outbound and inbound routes of the artificial ammunition equipment. The specific coverage of the key storage location nodes includes the storage door area, shelf aisle, and loading and unloading point. Using the key nodes of the warehouse entrance area as the visual starting point and the key nodes of the loading and unloading points as the visual ending point, visual image tracking is performed for the entire process of outbound operations. By swapping the positions of the visual starting point and the visual ending point, visual image tracking is performed for the entire process of inbound operations. The visual image tracking content corresponding to the entire outbound operation is as follows: Several outbound tracking nodes are set up between the visual start point and the visual end point. The outbound tracking nodes are numbered and denoted as i, i = 1, 2, 3, ..., n, where n is a natural number greater than 0. Several video recording devices are deployed at the visual start point, the visual end point and the outbound tracking nodes. The video recording devices are activated and capture several frames of video clips of the human-shaped ammunition equipment leaving the equipment warehouse. By connecting different image segments, an outbound scene animation is constructed. The outbound scene animation is used to represent the real-time image of the manned ammunition equipment at each position when the manned ammunition equipment is outbound. All real-time images are merged and then used as the image tracking result of the entire process corresponding to the outbound operation. The visual image tracking content corresponding to the entire process of the warehousing operation is as follows: Several inbound tracking nodes are set between the visual start point and the visual end point, and a robotic arm operation point is associated with each inbound tracking node. A robotic arm for controlling the artificial munition equipment is deployed at the robotic arm operation point. An image recording device is deployed at each inbound tracking node, and the corresponding imaging capture area of ​​each image recording device is set. The image recording device operates based on the imaging capture area. The image recording device obtains the image frame of the artificial munition equipment at each position within the imaging capture area. When the artificial munition equipment at any entry tracking node is not within the imaging capture area, the robotic arm of the robotic arm operation point associated with the current entry tracking node is manipulated to adjust the position of the artificial munition equipment so that it is within the imaging capture area. When it is within the imaging capture area, no operation is performed. Cargo transport robots are activated at the visual starting point and visual ending point. The cargo transport robots load the artificial munitions and equipment from the loading and unloading point and store the artificial munitions and equipment in the equipment warehouse. All the image frames of the artificial munitions and equipment in the warehouse are integrated and used as the image tracking results of the entire warehouse operation.

[0024] It should be further explained that, in the specific implementation process, the process of obtaining several frames of equipment orientation images based on the image tracking results, and obtaining entry and exit trace information based on all frames of equipment orientation images, includes: Set up a first processing window and a second processing window; The first processing window is used to process the image tracking results of the artificial ammunition equipment during the entire process of the outbound and inbound operations, and to preprocess the images of several frames of images of the artificial ammunition equipment being taken out of the equipment warehouse and several frames of images of the equipment being put into the equipment warehouse for storage. The image preprocessing includes image denoising, image enhancement, and layered distortion correction; The image denoising process is as follows: the first processing window selects a denoising area, which is a rectangular area composed of several pixel blocks. The pixel value of the pixel block at the center of the rectangular area is used as the pixel reference value. The pixel values ​​of other pixel blocks around the center of the rectangular area are unified as the pixel reference value. After adjusting the pixel values ​​of all pixel blocks, a standard noise image is obtained. The image enhancement process involves: setting several regions to be enhanced in a standard noisy image; detecting pixel contrast, light intensity, and color saturation in each region to be enhanced; obtaining any pixel block that does not meet the requirements for pixel contrast, light intensity, or color saturation; and marking it as a point to be enhanced under the region to be enhanced. Among them, pixel contrast, light intensity, and color saturation are used as enhancement type items; Group all points to be enhanced under the same enhancement type into a type point set, set several gradient enhancement intervals with different numerical ranges, and integrate the points to be enhanced under the same type point set into a gradient point cluster. Based on the gradient enhancement interval set according to the type point set corresponding to different enhancement type items, several different gradient point clusters are obtained, and pixel adjustment of several gradient point clusters is performed concurrently. The pixel adjustment includes increasing the pixel contrast value, increasing the illumination intensity value, and fitting the color saturation. After completing the pixel adjustment, the standard noise image is converted into a standard fitted image. It should be noted that several pixel blocks within the same gradient range share the common characteristic of having the same range of deviation values. Grouping them into a gradient point cluster and processing the image enhancement of all pixel blocks represented by the same gradient point cluster enables batch processing and effectively improves the processing efficiency of image enhancement.

[0025] The layered distortion correction involves dividing the standard fitted image into different focus plane layers based on the depth of field. Standard fitted images within the same focus plane layer are then integrated into a layered image set, which is used to obtain layered image sets corresponding to different focus plane layers. When the people, ammunition, or equipment being photographed are not in the same focus plane layer (i.e., at different distances from the camera), the same set of correction parameters cannot perfectly correct objects at all depths. Therefore, it is necessary to layer the images based on the depth of field values ​​to perform targeted distortion correction.

[0026] Set a sliding correction window, and use the sliding correction window to traverse each layer atlas. The sliding correction window covers and maps to each standard fitted image included in the layer atlas, and divides the standard fitted image into several sub-regions to be corrected. In each sub-region to be corrected, mark a number of distortion points and label the distortion type corresponding to each distortion point. The distortion types include radial distortion, tangential distortion, and shutter distortion. All distortion points of the same distortion type in all sub-regions to be corrected are grouped into a distortion set. Based on the number of distortion sets, a corresponding number of correction parameters are set. Each correction parameter is used to correct the distortion points under a corresponding distortion set, thereby converting the distortion points into normal points. After the sliding correction window completes the distortion correction of all different layered image sets, all standard fitted images are converted into the corresponding equipment frame interface image. The equipment frame interface image is covered with a complete image of human shadow ammunition equipment and the surrounding environment. The second processing window is used to reconstruct the equipment frame interface image, dividing the equipment frame interface image into several reconstruction areas of equal size, and sequentially identifying the image features under each reconstruction area. The image features include directional element features and non-directional element features. The directional element features are used to describe the outline appearance features of artificial ammunition equipment; The non-directional element features are used to describe the surrounding environmental features of artificial ammunition equipment; When a certain reconstruction region contains only directional element features, the current reconstruction region is retained as the object to be stitched. When a certain reconstruction region contains only non-directional element features, the current reconstruction region is removed. When a certain reconstruction region contains both directional and non-directional element features, the non-directional element features of the corresponding reconstruction region are stripped to obtain a reconstruction region containing only directional element features as the object to be stitched. All the objects to be stitched together are stitched together in chronological and positional order to construct several frames of equipment orientation images. Each frame of equipment orientation image is used to represent the real-time status of the manned ammunition equipment at an entry / exit location. In each frame of the equipment orientation image, select several trace calibration points. The trace calibration points include several contour boundary points, center points and corner points on the contour of the human shadow ammunition equipment. Connect the trace calibration points in the same equipment position in different frames of the equipment orientation image to obtain the entry and exit trace path. The entry and exit trace path is divided into several information nodes based on the operation timestamp. For each information node, the operation-related information of the artificial ammunition equipment at the current time and location is imported. The operation-related information includes equipment identity information, spatiotemporal trajectory information, context information and audit information. The spatiotemporal trajectory information describes when, where, and how the manned ammunition equipment moves. Specifically, it includes operation time, spatial location, event tags, and the ID of the associated camera. Event tags are used to record different operational behaviors of the equipment, such as picking it up from the shelf in the equipment warehouse, passing through the warehouse door, loading it onto a vehicle, and being scanned. These correspond to the concept of how the equipment moves. The associated camera ID indicates which camera captured the current equipment. The context information describes how all operations are completed and the surrounding environment during the operations. Specifically, it includes operator information, operation sequence, interactive device information, and environmental data. Operator information includes the operator ID (e.g., staff_007) associated with facial recognition or employee badge recognition, recording their activity range throughout the operation. The operation sequence records the specific actions of the personnel, including approaching the shelf, bending over, scanning equipment codes, lifting equipment, and moving equipment. The interactive device information involves relevant information about various devices used in the operation, such as the ID of the cargo transport robot, the ID of the handheld terminal, and the ID of the robotic arm. The environmental data records the real-time temperature and humidity inside the equipment warehouse during the entry and exit process. The audit information is used to describe whether there are any suspected anomalies in the operation behavior at different times and locations, and to retain evidence of suspected abnormal operation behavior. When an operation behavior is suspected of being abnormal, an anomaly flag is associated with the current information node, and a few seconds of video clips before and after the directional screen of the equipment corresponding to the current information node are stored in the cloud server path as evidence snapshot links. The operation-related information of all information nodes on the entry and exit trace path, along with the evidence snapshot links, are integrated into entry and exit trace information. This information is then encapsulated into a data packet, and the address of the data packet is mapped to a pre-deployed virtual machine.

[0027] It should be further explained that, in the specific implementation process, the process of analyzing the entry and exit trace information, extracting the exit feature set and the entry feature set, comparing them with the pre-configured entry and exit templates, and identifying abnormal entry and exit behaviors based on the comparison results includes: Several types of feature extraction rules are set up, and feature extraction is performed on the entry and exit trace information based on all feature extraction rules. After feature extraction is completed, different types of entry and exit related features are obtained. Entry and exit related features include operation context features, spatiotemporal trajectory features, equipment identity features, and behavior sequence features. The several types of feature extraction rules include the following: Rule 1: Quantity Consistency Rule; Rule 2: Path compliance rules; Rule 3: Identity Consistency Rule; Rule 4: Process Completeness Rule; The correspondence between different feature extraction rules and the corresponding inbound / outbound features is as follows: Rule 1: Quantity Consistency Rule—Contextual Features; Rule 2: Path Compliance Rule—Spatiotemporal Trajectory Characteristics; Rule 3: Identity Consistency Rule—Equipment Identity Characteristics; Rule 4: Process Completeness Rule—Behavioral Sequence Characteristics; Based on the entry and exit trace information, determine the operation type of the current artificial ammunition equipment. When the operation type is entry, extract all entry and exit related features related to the current entry to generate an entry feature set. When the operation type is exit, extract all entry and exit related features related to the current exit to generate an exit feature set. Pre-configure entry and exit templates. Each entry and exit template defines text recognition words that identify several entry and exit features. These text recognition words are used to characterize the behavioral features of the manned ammunition equipment during the correct execution of entry and exit. A rule engine is deployed, which connects to several feature extraction rules, inbound / outbound related features, and inbound / outbound templates. Then, it sequentially compares all operational behaviors of each manned ammunition equipment during the inbound / outbound process. When the inbound / outbound feature corresponding to the operational behavior fails to match the text recognition words of the inbound / outbound template, the operational behavior is filtered as an abnormal inbound / outbound behavior, until all abnormal inbound / outbound behaviors are filtered out.

[0028] It should be noted that the different feature extraction rules are used to extract the relevant features of the corresponding inbound and outbound operations as follows: The goal of the quantity consistency rule is to determine whether the planned inbound and outbound equipment quantity is consistent with the actual equipment quantity; the data source for the execution rule is: the fields of the business order number and the actual identified equipment list in the inbound and outbound trace information. The outbound demand table or inbound application table is queried based on the fields to obtain the calculated quantity. When the calculated quantity is inconsistent with the actual quantity, the quantity difference is used as the context feature. The objective of the path compliance rules is to determine whether the actual trajectory of equipment moving within the warehouse area conforms to the predetermined safe and efficient path. The data sources for executing the rules are: fields related to the "spatiotemporal trajectory sequence" in the entry and exit trace information; extracting all data points with location information (such as image coordinates, actual coordinates, and area labels) from the spatiotemporal trajectory sequence; performing a mandatory point check, which refers to key nodes that must be passed on the pre-set standard path, such as scanning points, weighing points, and exits; checking whether the fields of the spatiotemporal trajectory sequence cover all mandatory points; and then performing a restricted area check, which refers to pre-set prohibited areas (such as other shelf aisles); checking whether the fields of the spatiotemporal trajectory sequence include the restricted area. The goal of the identity consistency rule is to determine whether the equipment actually used is completely consistent with the equipment planned for use, and whether there has been any "switching" or "mistaken taking". The data source for executing the rule is: the fields in the inbound and outbound trace information that are associated with the business order number and the actual identified equipment list. Based on the business order number, the planned equipment ID set in the database is queried, and the actual identified equipment ID set is obtained. Then, set operations are performed to obtain different calculated sets. The computation set includes the following: Calculate the difference set 1: Equipment that is planned but not actually present, i.e., equipment missing; Calculate the difference set 2: equipment that exists in reality but is not planned, i.e., equipment surplus; Calculate the intersection: equipment that exists in both reality and the plan, i.e., equipment operation is correct; The computation set is used as the equipment identity feature of the shadow ammunition equipment; The goal of the process completeness rule is to determine whether an operation is strictly completed according to the pre-set standard operating procedure (SOP) and whether there are any missing steps or incorrect sequences. The data source for the execution rule is the fields corresponding to the event tag sequence and operation action sequence in the inbound and outbound trace information. The event tag sequence in the trajectory sequence and the operation action sequence in the behavior sequence are merged in chronological order to obtain a complete actual behavior sequence. The actual behavior sequence is compared with the pre-configured standard behavior sequence template through a sequence matching algorithm to calculate the missing steps and the incorrect step sequence, and the calculation result is used as the behavior sequence feature.

[0029] It should be further noted that in the specific implementation process, the process of extracting information from inbound and outbound behaviors and establishing a topological information tree for linked information push includes: Preset an information rule library, which stores rule forms and identification forms; Based on the information rule library, group the types and label the events of all inbound and outbound behaviors. The type grouping includes primary grouping and secondary grouping. The rule forms include preliminary screening forms and re-screening forms; The preliminary screening form is used for primary grouping of inbound and outbound behaviors. The preliminary screening form items record the behavior numbers of all abnormal inbound and outbound behaviors and the behavior numbers of normal inbound and outbound behaviors. All abnormal inbound and outbound behaviors are screened out through primary grouping, and the abnormal inbound and outbound behaviors are further grouped through the re-screening form to obtain the abnormal detail types of each abnormal inbound and outbound behavior, and no processing is done to normal inbound and outbound behaviors; The abnormal detail types recorded in the re-screening form include quantity abnormality, timing abnormality, and frequency abnormality. For example, when the inbound and outbound quantity of shadow ammunition equipment does not match the document, it is a quantity abnormality; when there is an outbound without an order in advance, or the inbound is not inspected by quality control, it is a timing abnormality; when the same shadow ammunition equipment performs inbound and outbound operations more than the preset number threshold within a time period, it indicates frequent inbound and outbound behaviors, corresponding to frequency abnormality; The identification form is used to label the events of abnormal inbound and outbound behaviors corresponding to different abnormal detail types, label each abnormal inbound and outbound behavior as a standardized abnormal structure event, and label event information. The event information includes event ID, abnormal type, timestamp, core entity, and event detail content; Among them, the event ID is the unique identifier of the standardized abnormal structure event, the abnormal type is quantity abnormality, timing abnormality, and frequency abnormality, the timestamp represents the specific time point when the abnormality occurs, the core entity is used to record the materials, personnel, equipment, and storage locations involved in the event, and the event detail content is used to record the specific content of the entire process of the event, such as the actual number of shadow ammunition equipment during inbound and outbound operations does not match the specified number recorded in the document.

[0030] Instantiate each standardized abnormal structure event into an event node, use the core entity of the standardized abnormal structure event as the mapping element of the event node, and establish event edges for mutual connection between event nodes. The event edges are used to represent the mutual relationship between the mapping elements in different event nodes. For example, person A belongs to handling group C, material B is stored at storage location D, and person E operates on equipment F; Connect all current event nodes and event edges based on the relationships between mapping elements to build an information subtree. Each information subtree represents all event information of a human shadow ammunition equipment entering and leaving the warehouse. Set a root node and connect all information subtrees to the root node to build a topological information tree. When a new standardized abnormal structure event is added, an information subtree corresponding to the current standardized abnormal structure event is established and the information subtree is mounted on the topology information tree. When performing an information retrieval of a certain artificial ammunition equipment entering or leaving the warehouse, the corresponding event ID and search terms are entered. The corresponding information subtree is retrieved from the topology information tree based on the event ID, and the corresponding event node on the information subtree is retrieved based on the search terms. The event information corresponding to the directly retrieved event nodes will be used as the direct demand information; Traverse all other event nodes on the event edge of the information subtree corresponding to the retrieved event node, take the event information of other event nodes as linkage information, divide the linkage information into depth levels, and fold the linkage information into several depth levels of data linkage layers. Among them, the event information of the event node is obtained from the corresponding information node in the virtual machine, and the operation-related information of the information node related to the event node and the evidence snapshot link are used together as the event information; Specifically, the event information of the event node corresponding to the event edge associated with the directly retrieved event node is taken as the linkage information with a depth of 1, and a data linkage layer with a depth of 1 is constructed to fold the corresponding linkage information. The event information of the event node on the event edge associated with the event node corresponding to the linkage information with a depth of 1 is taken as the linkage information with a depth of 2, and a data linkage layer with a depth of 2 is constructed to fold the linkage information with a depth of 2. This process continues until all event nodes in an information subtree have completed the construction of their corresponding data linkage layers. Based on the depth hierarchy, the folded linkage information within the data linkage layer is unfolded sequentially from low to high to push linkage information. The object performing the information retrieval decides whether to select the linkage information pushed from the corresponding data linkage layer, and makes decisions on abnormal inbound and outbound behaviors based on the directly retrieved information and linkage information.

[0031] Example 2, please refer to Figure 2 As shown, the present invention also provides an intelligent identification system for the entry and exit of manned ammunition equipment, the system comprising: The inbound / outbound module is used to register the inventory of each manned ammunition and equipment in the equipment warehouse, perform outbound operations based on the outbound demand table and inventory information, and perform inbound operations based on inventory status and inbound application information. The visual tracking module is used to perform visual image tracking on several pieces of artificial ammunition equipment entering and leaving the warehouse. Based on the image tracking results, several frames of equipment orientation images are obtained, and the entry and exit trace information is obtained based on all the equipment orientation images. The intelligent identification and information push module is used to analyze the entry and exit trace information, extract the exit feature set and the entry feature set, compare them with the pre-configured entry and exit templates, identify abnormal entry and exit behaviors based on the comparison results, extract information from the entry and exit behaviors, establish a topology information tree, and push related information.

[0032] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for intelligently identifying the entry and exit of human shadow ammunition equipment into a warehouse, characterized in that, The method comprises the following steps: Step S1: inventory registration is performed for each mannequin ammunition equipment in the equipment warehouse, warehouse-out operation is performed based on a warehouse-out demand table and inventory information, and warehouse-in operation is performed based on an inventory state and warehouse-in application information; Step S2: visual image tracking is performed on a plurality of mannequin ammunition equipments in and out of the warehouse, a plurality of frames of equipment orientation pictures are obtained based on the image tracking results, and warehouse-in and warehouse-out trace information is obtained according to the equipment orientation pictures of all frames; Step S3: the warehouse-in and warehouse-out trace information is analyzed, warehouse-out feature sets and warehouse-in feature sets are extracted, and comparison is performed with preconfigured warehouse-in and warehouse-out templates, abnormal warehouse-in and warehouse-out behaviors are identified based on the comparison results, information extraction is performed on the warehouse-in and warehouse-out behaviors, a topological information tree is established, and linkage information is pushed.

2. The method of claim 1, wherein the method is characterized by: The process of inventory registration for each mannequin ammunition equipment in the equipment warehouse, warehouse-out operation based on a warehouse-out demand table and inventory information, and warehouse-in operation based on an inventory state and warehouse-in application information comprises: statistical mannequin ammunition equipment equipment identity information creates respective registration texts, all registration texts are stored in a database, the registration texts are audited, and the inventory registration of the mannequin ammunition equipment and the rejection processing of the registration texts are completed based on the audit results; a warehouse-out demand table records the mannequin ammunition equipment required to be taken out, data monitoring is performed on all registration texts stored in the database and completed the audit, the warehouse-in and warehouse-out records of the mannequin ammunition equipment under each equipment area are obtained, and the inventory information is obtained based on the warehouse-in and warehouse-out records; the mannequin ammunition equipment required to be taken out is selected from the equipment warehouse according to the warehouse-out demand table, the whole process of the warehouse-out operation is supervised, the mannequin ammunition equipment is transported to the equipment warehouse according to the warehouse-in demand table, and the whole process of the warehouse-in operation is supervised.

3. The method of claim 2, wherein the method further comprises: determining whether the person is a person of interest based on the determined face feature information and the determined body feature information. The process of visual image tracking of a plurality of mannequin ammunition equipments in and out of the warehouse comprises: a plurality of warehouse position key nodes are set on the warehouse-out path channel and the warehouse-in path channel, the coverage range includes a warehouse door area, a shelf channel and a loading and unloading point, the warehouse door area is taken as a visual starting point, the loading and unloading point is taken as a visual ending point, visual image tracking is performed on the whole process of the warehouse-out operation, the positions of the visual starting point and the visual ending point are exchanged, and visual image tracking is performed on the whole process of the warehouse-in operation; a plurality of warehouse-out tracking nodes are set, a plurality of image recording devices are arranged at the visual starting point, the visual ending point and the warehouse-out tracking nodes, a plurality of frames of image segments of the warehouse-out operation are photographed by the image recording devices, the warehouse-out scene animation is obtained by splicing different image segments as the image tracking result of the warehouse-out operation; a plurality of warehouse-in tracking nodes are set, a mechanical arm operation point corresponding to the warehouse-in tracking node is set for deploying the mechanical arm, an image recording device is arranged at the warehouse-in tracking node, an imaging interception area is set to obtain the warehouse-in image frames of the mannequin ammunition equipment at each position, when the mannequin ammunition equipment at any warehouse-in tracking node is not in the imaging interception area, the position of the mannequin ammunition equipment at the current warehouse-in tracking node is adjusted to the imaging interception area by the mechanical arm at the current warehouse-in tracking node; the mannequin ammunition equipment is loaded from the loading and unloading point and stored in the equipment warehouse, all warehouse-in image frames of the whole process of the warehouse-in operation are integrated as the image tracking result of the warehouse-in operation.

4. The method of claim 3, wherein the method further comprises: determining whether the person is a person of interest based on the determined face feature information and the determined body feature information. The process of obtaining the several frames of equipment directional pictures based on the image tracking results comprises: setting a first processing window and a second processing window; the first processing window is used for image preprocessing of the several frames of image segments and the warehouse entry image frame, including image denoising, image enhancement and hierarchical distortion correction; a standard noise image is obtained through image denoising, the standard noise image is converted into a standard fitting image through image enhancement, the standard fitting image is subjected to hierarchical distortion correction to obtain an equipment frame interface picture, and the equipment frame interface picture covers a complete human shadow ammunition equipment picture and a surrounding environment picture; the second processing window is used for picture reconstruction of the equipment frame interface picture, the equipment frame interface picture is divided into several reconstruction regions, and image features under each reconstruction region are identified, the image features including directional element features and non-directional element features; when there is only directional element feature in a certain reconstruction region, the reconstruction region is retained as a splicing object, when there is only non-directional element feature, the reconstruction region is removed, and when there are both directional element features and non-directional element features, the reconstruction region is subjected to non-directional element feature stripping, and the reconstruction region including only directional element features is taken as a splicing object; all the splicing objects are spliced in time sequence and position sequence to obtain the several frames of equipment directional pictures.

5. The human shadow ammunition equipment warehouse in and out intelligent identification method according to claim 4, characterized in that, The process of obtaining the warehouse entry and exit trace information based on the equipment directional pictures of the total number of frames comprises: selecting several trace calibration points in each frame of equipment directional picture, connecting the trace calibration points at the same device position on different frames of equipment directional picture to obtain a warehouse entry and exit trace path, and dividing the warehouse entry and exit trace path into several information nodes based on operation time stamps; importing operation related information of the human shadow ammunition equipment at the current time and position for each information node, linking and integrating the operation related information and evidence snapshots of all the information nodes on the warehouse entry and exit trace path as warehouse entry and exit trace information, encapsulating the warehouse entry and exit trace information as a data packet, and mapping the address of the data packet to a pre-deployed virtual machine.

6. The human shadow ammunition equipment warehouse in and out intelligent identification method according to claim 5, characterized in that, The process of image enhancement and hierarchical distortion correction comprises: setting several to-be-enhanced regions in the standard noise image, taking pixel contrast, illumination intensity and color saturation as enhancement type items, obtaining a pixel block that does not meet the requirements of any enhancement type item, and identifying it as a to-be-enhanced point in the to-be-enhanced region; grouping all the to-be-enhanced point positions under the same enhancement type item into a type point set, setting several gradient enhancement intervals, integrating the to-be-enhanced point positions under the same gradient enhancement interval in the same type point set into a gradient point cluster, performing pixel adjustment of several gradient point clusters to convert the standard noise image into a standard fitting image; dividing the standard fitting image into different focus plane layers based on the depth of field, integrating the standard fitting images of the same focus plane layer into a hierarchical atlas, setting a sliding correction window to perform window traversal on each hierarchical atlas, and dividing the standard fitting image into several to-be-corrected sub-regions; Mark a number of distortion points in each to be corrected sub-region, and mark the distortion type of the distortion points, classify the distortion points of the same distortion type into a distortion set, set a corresponding number of correction parameters based on the number of distortion sets, each correction parameter is used for correcting the distortion points in the corresponding distortion set, and then converting the distortion points into normal points, after completing the distortion correction of all layered atlas, converting all standard fitting images into equipment frame interface images.

7. The method of claim 6, wherein the method further comprises: determining whether the person is a person of interest based on the determined face feature information and the determined body feature information. In the specific implementation process, the in-out warehouse trace information is analyzed, the out-of-warehouse feature set and the in-warehouse feature set are extracted, and the pre-configured in-out warehouse template is compared. The process of identifying abnormal in-out warehouse behavior based on the comparison result includes: A plurality of types of feature extraction rules are set to extract features from the in-out warehouse trace information to obtain different types of in-out warehouse related features, including operation context features, space-time trajectory features, equipment identity features, and behavior sequence features; Based on the in-out warehouse trace information, the operation type of the current human shadow ammunition equipment is determined. When the operation type is in-warehouse, the in-out warehouse related features related to the current in-warehouse are extracted to generate an in-warehouse feature set. When the operation type is out-of-warehouse, the in-out warehouse related features related to the current out-of-warehouse are extracted to generate an out-of-warehouse feature set; The in-out warehouse template is pre-configured, the in-out warehouse template includes a plurality of text identification words, the rule engine is deployed to access a plurality of feature extraction rules, in-out warehouse related features and in-out warehouse template, and the operation behavior of each human shadow ammunition equipment in the in-out warehouse process is compared in turn. When the in-out warehouse features of the operation behavior are not successfully compared with the text identification words of the in-out warehouse template, the operation behavior is screened as an abnormal in-out warehouse behavior, and all abnormal in-out warehouse behaviors are screened.

8. The method of claim 7, wherein the method further comprises: determining whether the person is a person of interest based on the determined face feature information and the determined body feature information. The process of information extraction on in-out warehouse behavior includes: The preset information rule library is used to store rule forms and identification forms; Based on the information rule library, all in-out warehouse behaviors are grouped by type and event labeled. Type grouping includes primary grouping and secondary grouping. Rule forms include preliminary screening forms and rescreening forms; The preliminary screening form is used for primary grouping of in-out warehouse behaviors to screen all abnormal in-out warehouse behaviors. The rescreening form performs secondary grouping on the abnormal in-out warehouse behaviors to obtain the abnormal detail type of each abnormal in-out warehouse behavior. Normal in-out warehouse behaviors are not processed. Abnormal detail types include quantity abnormalities, time sequence abnormalities, and frequency abnormalities; The identification form is used to event label abnormal in-out warehouse behaviors of different abnormal detail types, label each abnormal in-out warehouse behavior as a standardized abnormal structure event, and label event information. Event information includes event ID, abnormal type, timestamp, core entity, and event detail content.

9. The intelligent identification method for the entry and exit of manned ammunition equipment according to claim 8, characterized in that, The process of establishing a topology information tree for linked information pushing includes: Connect all event nodes and event edges based on the relationship of mapping elements, and then build an information subtree. Set the root node and connect all information subtrees to the root node to get the topology information tree. When a new standardized abnormal structure event is added, the information subtree corresponding to the current standardized abnormal structure event is established and mounted to the topology information tree. When the information retrieval of a certain in-out warehouse mannequin ammunition equipment is performed, the corresponding event ID and retrieval word are input, the corresponding information subtree is retrieved from the topological information tree based on the event ID, and the corresponding event node on the information subtree is retrieved based on the retrieval word; The event information of the directly retrieved event node is taken as the direct demand information; All other event nodes on the event edge of the information subtree corresponding to the retrieved event node are traversed, the event information of the other event nodes is taken as the linkage information, the linkage information is divided into several depth levels, and the linkage information is folded into several depth levels of data linkage layers; The folded linkage information in the data linkage layers is unfolded in the order from low to high based on the depth level, whether to pick the linkage information in the data linkage layers is determined by the object of the information retrieval, the direct retrieval information and the linkage information are used to perform decision processing on the abnormal in-out warehouse behavior.

10. A human shadow ammunition equipment warehouse in and out intelligent identification system for implementing the human shadow ammunition equipment warehouse in and out intelligent identification method of any one of claims 1 to 9. The system comprises: An in-out warehouse module for inventory registration of each mannequin ammunition equipment in the equipment warehouse, out-of-warehouse operation based on the out-of-warehouse demand table and the inventory information, and in-warehouse operation based on the inventory state and the in-warehouse application information; A visual tracking module for visual image tracking of a plurality of mannequin ammunition equipment in the in-out warehouse, obtaining a plurality of equipment orientation pictures based on the image tracking results, and obtaining in-out warehouse trace information according to the equipment orientation pictures of all frames; An intelligent identification and information pushing module for analyzing the in-out warehouse trace information, extracting the out-of-warehouse feature set and the in-warehouse feature set, comparing with the pre-configured in-out warehouse template, identifying the abnormal in-out warehouse behavior based on the comparison result, extracting the in-out warehouse behavior information, establishing a topological information tree, and pushing the linkage information.

Images