Intelligent storage digital visual management system and method based on Internet of Things

Abstract

The invention relates to the technical field of Internet of Things perception, in particular to an intelligent storage digital visual management system and method based on the Internet of Things. The method comprises the following steps: acquiring intelligent storage structure data; constructing a storage three-dimensional model according to the intelligent storage structure data; carrying out material carrying simulation based on the storage three-dimensional model to obtain material carrying data; identifying a high-frequency congestion area based on the material handling data; carrying out cargo placement analysis according to the high-frequency congestion area to obtain cargo placement data; identifying a cargo spatial form based on the cargo placement data to obtain cargo spatial form data; carrying out cargo gravity center shift detection according to the cargo spatial form data to obtain cargo gravity center shift data; and according to the cargo gravity center shift data, acquiring a cargo unique label to obtain cargo label data. The goods stacking safety rate and the goods shelf structure operation stability rate in intelligent storage are improved based on the Internet of Things sensing technology.

Description

Technical Field

[0001] The present invention relates to the field of Internet of Things perception technology, and in particular to an intelligent warehousing digital visualization management system and method based on the Internet of Things. Background Art

[0002] Traditional intelligent warehousing digital visualization management has many defects when dealing with highly dynamic, multi-deformation, and multi-material cargo stacking scenarios. For example, most existing systems only rely on static warehouse layout data and standard material labels for visualization modeling, and lack the ability to perceive real-time handling paths, high-frequency transportation areas, and dynamic changes in the center of gravity of cargo, resulting in delayed cargo placement optimization and risk warning responses. For goods with wrinkles, reflective materials, and transparent packaging, the depth information recognition accuracy is low, making it difficult for the system to accurately reconstruct the cargo space form and volume model, affecting the accuracy of subsequent shelf load assessments. It is mostly based on fixed thresholds to determine whether the shelf is overloaded, and does not combine the cargo weight, volume, structural stacking conditions, and actual shelf load history data for fatigue analysis. It is difficult to achieve predictive management of shelf connection node fatigue, structural deformation, and service life, which ultimately leads to insufficient accuracy of the overall warehouse visualization management system and delayed warnings, which cannot meet the current high-density and frequently dynamically changing intelligent warehousing scenarios. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide an intelligent warehouse digital visualization management system and method based on the Internet of Things to solve at least one of the above technical problems.

[0004] To achieve the above objectives, a method for intelligent warehouse digital visualization management based on the Internet of Things includes the following steps:

[0005] Step S1: Acquire intelligent warehouse structure data; construct a three-dimensional warehouse model based on the intelligent warehouse structure data; perform material handling simulation based on the three-dimensional warehouse model to obtain material handling data;

[0006] Step S2: Identify high-frequency congestion areas based on the material handling data; perform cargo placement analysis based on the high-frequency congestion areas to obtain cargo placement data; identify cargo spatial form based on the cargo placement data to obtain cargo spatial form data; perform cargo center of gravity offset detection based on the cargo spatial form data to obtain cargo center of gravity offset data;

[0007] Step S3: Collect unique cargo labels based on cargo center of gravity offset data to obtain cargo label data; upload cargo label data to the warehouse management system, and display cargo storage status to obtain a real-time warehouse status diagram; perform shelf overload anomaly analysis based on the real-time warehouse status diagram to obtain shelf overload anomaly data;

[0008] Step S4: Perform shelf structure fatigue detection based on the shelf overload abnormality data to obtain shelf structure fatigue data; predict shelf service life based on the shelf structure fatigue data to obtain shelf life data; and generate a shelf health heat map based on the shelf life data.

[0009] The present invention uses traditional intelligent warehouse digital visualization management to deal with many defects in dealing with highly dynamic, multi-deformation, and multi-material cargo stacking scenarios. For example, most existing systems only rely on static warehouse layout data and standard material labels for visualization modeling, and lack the ability to perceive real-time handling paths, high-frequency transportation areas, and dynamic changes in the center of gravity of cargo, resulting in delayed cargo placement optimization and risk warning response. For goods with wrinkles, reflective, and transparent packaging, the depth information recognition accuracy is low, making it difficult for the system to accurately reconstruct the cargo space form and volume model, affecting the accuracy of subsequent shelf load assessments. It mostly judges whether the shelf is overloaded based on a fixed threshold, and does not combine the cargo weight, volume, structural stacking conditions, and actual shelf load history data for fatigue analysis. It is difficult to achieve predictive management of shelf connection node fatigue, structural deformation, and service life, which ultimately leads to insufficient accuracy of the overall warehouse visualization management of the system and delayed warnings, which cannot meet the current high-density and frequently dynamically changing intelligent warehousing scenario requirements.

[0010] Preferably, this specification also provides an Internet of Things-based intelligent warehouse digital visualization management system, which is used to execute the above-mentioned Internet of Things-based intelligent warehouse digital visualization management method, and the Internet of Things-based intelligent warehouse digital visualization management system includes:

[0011] The material handling simulation module is used to obtain intelligent warehouse structure data; build a three-dimensional warehouse model based on the intelligent warehouse structure data; and perform material handling simulation based on the three-dimensional warehouse model to obtain material handling data;

[0012] The cargo center of gravity offset detection module is used to identify high-frequency congestion areas based on material handling data; perform cargo placement analysis based on the high-frequency congestion areas to obtain cargo placement data; identify cargo spatial form based on the cargo placement data to obtain cargo spatial form data; and perform cargo center of gravity offset detection based on the cargo spatial form data to obtain cargo center of gravity offset data.

[0013] The shelf overload anomaly analysis module is used to collect unique cargo labels based on cargo center of gravity offset data to obtain cargo label data; upload cargo label data to the warehouse management system and display the cargo storage status to obtain a real-time warehouse status diagram; perform shelf overload anomaly analysis based on the real-time warehouse status diagram to obtain shelf overload anomaly data;

[0014] The shelf life prediction module is used to perform shelf structure fatigue detection based on shelf overload abnormal data to obtain shelf structure fatigue data; predict shelf life based on shelf structure fatigue data to obtain shelf life data; and generate a shelf health heat map based on the shelf life data.

[0015] The intelligent warehouse digital visualization management system based on the Internet of Things of the present invention can implement any one of the intelligent warehouse digital visualization management methods based on the Internet of Things of the present invention, and is used to combine the operation and signal transmission medium between various modules to complete the intelligent warehouse digital visualization management method based on the Internet of Things. The internal modules of the system cooperate with each other to improve the safety rate of cargo stacking and the stability rate of shelf structure operation in intelligent warehousing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0017] Figure 1 This is a schematic diagram of the steps of an intelligent warehouse digital visualization management method based on the Internet of Things of the present invention;

[0018] Figure 2 Detailed step flow diagram of step S1 in the present invention;

[0019] Figure 3 Detailed flowchart of step S16 in the present invention;

[0020] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0021] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are 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 those skilled in the art without making any creative work are within the scope of protection of the present invention.

[0022] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0023] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0024] To achieve this, please refer to Figures 1 to 3 The present invention provides an intelligent warehouse digital visualization management method based on the Internet of Things, the method comprising the following steps:

[0025] Step S1: Acquire intelligent warehouse structure data; construct a three-dimensional warehouse model based on the intelligent warehouse structure data; perform material handling simulation based on the three-dimensional warehouse model to obtain material handling data;

[0026] In this embodiment, a combination of laser radar (LiDAR) and millimeter-wave radar is used to scan the warehouse structure. The LiD AR model used is Velodyne HDL-32E, with a scanning accuracy of ±2cm, a point cloud resolution of 700,000 points / second, and a scanning angle covering 360° horizontal and 40° vertical ranges. The structure is completed in combination with the data of the millimeter-wave radar. The millimeter-wave radar model is Texas Instruments AWR1843, with a detection range of 0.1m to 100m and an accuracy of ±5cm. All structural data are synchronously filtered, denoised, and converted to a coordinate system through the edge computing device Jetson AGX Orin. The storage space is spliced ​​into a unified coordinate system using a point cloud registration method based on the ICP algorithm (iterative nearest point). The data point density after coordinate unification is controlled to be more than 1,000 points per square meter. During the 3D model construction process, the original point cloud was input into MeshLab software and a mesh model was generated using the Poiss on Surface Reconstruction algorithm. Ground obstacles less than 0.5m in height were marked as low-level obstacles and designated as inaccessible areas. When modeling the shelves, structural connections were annotated based on the shelf material properties (steel structure) and node connection method, with a modeling accuracy of 0.01m. After the model was completed, the material handling path simulation was performed using a self-developed material handling trajectory simulation module. Parameters were set to a handling device size of 1.2m × 0.8m, a minimum turning radius of 1.5m, a speed of 0.8m / s, an 8-hour simulation time span, and a path recording frequency of 1Hz. Handling data included the handling device's position (X, Y, and Z coordinates) in the 3D model at each moment, its direction vector (represented by quaternions), path length, weight handled per unit time, and the number of rest stops.

[0027] Step S2: Identify high-frequency congestion areas based on the material handling data; perform cargo placement analysis based on the high-frequency congestion areas to obtain cargo placement data; identify cargo spatial form based on the cargo placement data to obtain cargo spatial form data; perform cargo center of gravity offset detection based on the cargo spatial form data to obtain cargo center of gravity offset data;

[0028] In this example, during the material handling data analysis, a path frequency statistics method based on grid thermal clustering was used to divide the warehouse space into 0.5m×0.5m×0.5m cubic grid cells. For each cell, the frequency of traffic flow per unit time was counted. The threshold for high-frequency congestion areas was set to the upper quartile (Q3) of the warehouse's average daily traffic flow rate + 1.5 × the interquartile range (IQR). Areas with unusually high traffic flow were identified as high-frequency congestion areas at a 95% confidence level. During the cargo placement analysis phase, an RGB-D camera (Intel RealSense D455) was used to capture the three-dimensional structure of the cargo. Point clouds for each item were reconstructed using the OpenCV and PCL libraries. The maximum bounding box dimensions of the cargo were extracted, and the dimensional parameters were recorded: length L (0.2–1.2m), width W (0.2–0.8m), and height H (0.1–1.0m). Spatial morphological data was stored as a six-dimensional structure matrix: [L, W, H, aspect ratio, surface-to-volume ratio, center of gravity]. During the gravity center shift detection, a force sensor (Tekscan 5330N) was used to collect the bottom pressure distribution of the goods during stacking or handling. The bottom surface was divided into 5×5 grid areas, with a single grid area of ​​64cm. 2 The support center of gravity is calculated using the force data from each grid. Based on the Euclidean distance between the cargo's own center of gravity (calculated using 3D morphology simulation) and the support center of gravity, any offset greater than 15% of the shortest side length is flagged as a shift. The typical threshold for shift determination is ≥0.08m. A triaxial accelerometer is also used to detect cargo tilt. If the tilt angle is greater than 5° and the offset is directed toward the support edge, a center of gravity shift event is recorded.

[0029] Step S3: Collect unique cargo labels based on cargo center of gravity offset data to obtain cargo label data; upload cargo label data to the warehouse management system, and display cargo storage status to obtain a real-time warehouse status diagram; perform shelf overload anomaly analysis based on the real-time warehouse status diagram to obtain shelf overload anomaly data;

[0030] In this embodiment, each piece of cargo with a center of gravity offset is scanned with its corresponding RFID tag. The tag model is Alien ALN-9640, the tag protocol is EPC Class 1 Gen 2, the read / write distance is set to 4m, and a UHF band reader (Impinj R700 series) is used to obtain a unique EPC code. The image captured by the camera is bound to the warehouse coordinate system. Each tag data includes: [EPC code, center of gravity offset (m), detection time, shelf number, 3D coordinates (x, y, z)] All tag data is transmitted to the upper-level warehouse management server via the warehouse's on-site LoRa wireless communication gateway. The real-time warehouse status diagram is displayed using a combination of a 2D orthogonal expansion view and a 3D dynamic heat map. The shelf status is dynamically loaded on the web using the CesiumJS graphics library. The label color of each piece of cargo is set according to the center of gravity offset: green for an offset <0.05m, orange for 0.05–0.10m, and red for >0.10m. In overload anomaly analysis, the maximum load of a single-layer shelf is set to 300kg. The mass attribute is bound to each item when it enters the warehouse. If the total mass of any single-layer cargo on the shelf exceeds the set value, it will be marked as a red abnormal area through the management end. The real-time reporting data includes: [shelf number, current load mass, maximum allowable mass, overload amount].

[0031] Step S4: Perform shelf structure fatigue detection based on the shelf overload abnormality data to obtain shelf structure fatigue data; predict shelf service life based on the shelf structure fatigue data to obtain shelf life data; and generate a shelf health heat map based on the shelf life data.

[0032] In this embodiment, the fatigue detection of the shelf structure adopts the strain gauge array layout method. Four strain gauges (model KYOWA KFG-5-120-C1-11) are pasted at the horizontal connection of each group of shelves. The strain measurement accuracy is ±5με and the sampling frequency is 20Hz. The temperature and humidity environment (DHT22 sensor, error ±0.5℃, ±2%RH) are recorded synchronously to correct the strain drift. The historical overload data and the real-time strain value are used to calculate the cumulative load times. The Palmgren-Miner linear cumulative damage theory is used to calculate the fatigue accumulation factor C=Σ(n / N), where n is the actual load times and N is the number of cycles allowed at this stress level; when C≥1.0, it is determined that the structure has reached the fatigue limit. The shelf life prediction is based on the SN curve (stress-life curve) database. The matching shelf material is Q235 steel with a yield strength of 235MPa. The expected remaining service life is calculated according to the measured maximum stress peak. The typical setting is a target life of 10 6 cycles. If the measured stress is 80 MPa, the permissible number of cycles is about 5×10 5The remaining lifespan is estimated based on the current number of cycles and output as a time dimension (average handling cycle converted to days). The shelf health heat map uses shelf number as an index and uses color to distinguish health levels (green for lifespan >90 days, yellow for 30–90 days, and red for lifespan <30 days). A WebGL rendering engine is used for 3D display, and the warehouse management terminal simultaneously displays the lifespan layer, which is linked to the handling task scheduling logic to avoid delivering heavy goods to red areas.

[0033] Preferably, step S1 is specifically as follows:

[0034] Step S11: Acquire intelligent warehousing structure data;

[0035] In this embodiment, a high-precision LiDAR and panoramic camera system are deployed to scan and capture the internal structure of the warehouse. The LiDAR device used is a RIEGL VZ-400i, with a scanning distance accuracy of ±5mm, a horizontal field of view of 360°, a vertical field of view of 100°, a resolution of 0.002°, and a point cloud sampling frequency of 500,000 points / second. The LiDAR data is synchronized with the pose information output by the IMU (Inertial Navigation Unit), and the global coordinates are unified using a GNSS-assisted positioning system to ensure that the error does not exceed 2cm. The panoramic image acquisition device is a Matterport Pro2, with a resolution of 134MP and a collection interval of 3m, covering the entire warehouse corridor and shelf area. After the structural data is collected, the point cloud data is fused with the image data to extract the component outline. The unified output format is PCD (point cloud data) and EXIF ​​(image data) files, and saved to the edge server cache for subsequent building component identification.

[0036] Step S12: Identify building components based on the intelligent storage structure data to obtain building component data;

[0037] In this embodiment, building component recognition adopts a rule-driven and shape semantic fusion strategy. The point cloud data obtained in S11 is processed in a layered slice format, with the layer height interval set to 0.5m. Each layer of slice is extracted using the RANSAC algorithm (random sampling consistency) to identify structures such as walls, floors, and floor slabs; for example, continuous vertical planes with an area greater than 4m 2The plane with an inclination angle of <5° is marked as the ground or platform. For the shelf components, the DBSCAN density clustering algorithm is used to extract the strip-like dense structure, setting the minimum number of sample points to 30 and the ε neighborhood radius to 0.15m to identify steel columns and beams. The image data is parallel labeled with component types using YOLOv5s (input resolution of 640×640), and the results are fused with the point cloud labels for IoU matching verification (threshold ≥ 0.7) to complete the confirmation. The final output of the building component data is a structured JSON file with the following fields: component type (wall, column, floor, etc.), 3D bounding box coordinates (Xmin, Ymin, Zmin, Xmax, Ymax, Zmax), material category (steel, concrete, etc.), and relative spatial position index.

[0038] Step S13: Divide the space units based on the building component data to obtain storage space partition data;

[0039] In this embodiment, the spatial unit division is based on the component data and is processed by regular gridding. Taking the ground area as the reference plane, the entire storage area is divided into spatial cube units with a length of 1.5m × a width of 1.5m × a height of 2.0m. Before division, a Boolean operation is performed on the component boundaries to eliminate the areas that overlap with components such as walls and closed beams; if the overlapping volume ratio of the component boundaries within the unit body is greater than 30%, the unit is set as a prohibited area. A spatial adjacency graph is established for the available spatial units, and the accessible boundaries and connection directions of each spatial unit are recorded, and a six-neighborhood connection strategy is adopted: up, down, left, right, front, and back. The spatial unit is named with a coordinate index (for example: S_03_02_01 represents a unit with X=3, Y=2, and Z=1), and the functional information such as the shelf area, loading and unloading area, and passageway is marked. Finally, all valid units are output as storage space partition data in the format of a three-dimensional index matrix + function identification table.

[0040] Step S14: Reconstructing the cargo space layout according to the storage space partition data to obtain cargo space layout data;

[0041] In this embodiment, the cargo location layout reconstruction is mapped based on the matching relationship between the spatial partition data and the shelf component structure. First, the area containing the cross structure of steel columns and beams in the spatial unit is extracted and identified as the shelf volume space, the shelf size range is recorded (for example, the standard pallet shelf size is 1.2m×1.0m×1.5m), and its direction vector and shelf number are determined. All shelf areas are layered, and the height of each layer of cargo + safety distance (default 0.3m) is used as the height increment unit to reconstruct the cargo location structure of each layer. Each cargo location record parameter includes: cargo location number (such as H_01_03_L2 represents the 2nd layer of the 3rd column of shelf No. 1), spatial location index (to which it belongs, spatial unit ID), maximum allowable load capacity (standard value is 200kg), cargo location size range (length, width and height) and access channel direction (forward / side entry / backward entry). The cargo location layout data is output in the form of a structured table and spatial index, and a cargo location space mapping table bound to the three-dimensional coordinate system is generated at the same time (for subsequent three-dimensional model assembly and path planning module call).

[0042] Step S15: Modeling the structural hierarchical relationship based on the cargo location layout data to obtain warehouse structure diagram metadata; assembling a three-dimensional space model based on the warehouse structure diagram metadata to obtain a three-dimensional warehouse model;

[0043] In this embodiment, the structural hierarchical relationship modeling constructs a warehouse hierarchical topology map based on the cargo location layout. The top-level node is the warehouse number, which is divided into regional nodes (such as the receiving area, shelf area, return area, etc.). Each regional node is mounted with a shelf node, and the shelf node is further refined into the cargo location node. The hierarchical relationship is stored in the form of an adjacency matrix to represent the affiliation and connection between each node. The topological map data structure is stored in an adjacency linked list to save space. The metadata includes: node number, node type, parent node ID, spatial coordinates, and associated component ID. During the assembly of the three-dimensional spatial model, the cargo location element is bound to the component three-dimensional bounding box data and assembled in sequence according to the hierarchical topological structure. The assembly accuracy error is controlled within 5mm. Spatial collision detection is performed on each element when assembling (using the AABB bounding box detection algorithm) to ensure that there is no overlap and conflict. The final generated three-dimensional model format is in the glTF 2.0 standard, which is convenient for Web visualization engines to call. Each element in the model is bound to a unique ID and can be directly referenced in subsequent handling simulation, path tracing, and visual inspection.

[0044] Step S16: Perform material handling simulation based on the three-dimensional warehouse model to obtain material handling data.

[0045] In this embodiment, the material handling simulation is based on a three-dimensional warehouse model to calculate the path and model the movement of the handling equipment. The modeling parameters of the handling equipment include: equipment type (e.g., forklift), size (1.2m×0.8m), minimum turning radius (1.6m), maximum handling weight (1,000kg), maximum speed (1.5m / s), acceleration limit (0.5m / s), and the like. 2 ). The three-dimensional coordinates of the starting point and end point of the transportation are extracted from the goods entry point and the designated cargo location respectively. The path planning adopts a spatial navigation system, and the spatial partition grid is used as a node graph. The path cost function combines the distance, direction switching cost and congestion penalty weight, where the weight parameters are set as: basic cost = 1, direction switching penalty = 0.3, and per-unit congestion frequency cost = 0.5. After the path is generated, the position (X, Y, Z and direction) of the transportation equipment is recorded at a time interval of 1 Hz to generate complete transportation trajectory data. The transportation data contains the following fields: [timestamp, equipment ID, coordinate position, current cargo ID, path length, speed, whether to stop]. Each task is fully recorded and saved to the IoT data server for subsequent analysis and scheduling optimization.

[0046] Preferably, step S16 is specifically as follows:

[0047] Step S161: performing storage path gridding processing based on the three-dimensional warehouse model to obtain path grid data;

[0048] In this embodiment, based on the three-dimensional warehouse model in glTF format, the voxel partitioning method is first used to perform path gridding processing on the entire warehouse space. A uniform cubic grid with a Voxel Size set to 0.5m×0.5m×0.5m is used to divide the X, Y, and Z axes into equal intervals. Each voxel block is marked as occupied according to whether there is a physical structure. If the Euclidean distance between the center point of the voxel and the boundary of the component is less than 0.1m, the voxel is marked as "inaccessible"; otherwise, it is marked as "accessible". The component boundary information is obtained by parsing the bounding box field (Xmin, Ymin, Zmin, Xmax, Ymax, Zmax) in the warehouse structure graphic element. The gridding process adopts an eight-adjacent connection method, that is, each passable voxel records the connectivity status with its eight adjacent voxels and stores it as a sparse adjacency matrix structure (using COO format). The path grid data field includes: voxel ID (three-dimensional index value), occupancy status, adjacency index list, and spatial coordinate range. The path grid data is ultimately stored in a local cache, and an API interface is provided for subsequent path identification and planning.

[0049] Step S162: Identify the material transport channel according to the path grid data to obtain transport channel data;

[0050] In this embodiment, based on the path grid data, the Flood Fill algorithm is used to detect the channel connectivity domain. All "accessible" voxels are used as seed points, and the region is expanded according to the six-sided connection principle to form a continuous accessible area. The maximum length, width, height dimensions and minimum channel width of each connected area are calculated; if the minimum channel width is ≥1.2m and the height is ≥1.8m (meeting the standard forklift traffic requirements), the connected area is determined to be a "transportable channel". The transport channel data is expressed in the form of a two-dimensional path skeleton. The central path segment is obtained through Skeletonization processing, and each segment is represented by a direction vector (unit vector), starting and end point coordinates, and a path width field. Each transport channel records the spatial block number to which it belongs, the associated shelf ID, and the unique channel number. Finally, the transport channel data is output in a graph structure, each node represents a channel intersection, and each edge represents a drivable segment.

[0051] Step S163: Performing an AGV reachability analysis based on the transport channel data to obtain reachable path data;

[0052] In this embodiment, the accessibility analysis is based on the graph structure composed of the transport channel data, and the Dijkstra shortest path algorithm is used for path search. First, the basic parameters of the automatic guided vehicle (AGV) are imported: the minimum turning radius is 1.2m, the maximum slope tolerance angle is 8°, and the minimum channel width requirement is 1.0m. The geometric structure of the transport channel is calculated by checking the continuity of the line segment and the turning angle to calculate the curvature radius; if the curvature radius of any turning point is less than the minimum turning radius of the AGV, the channel section is marked as "inaccessible". After removing all path segments that do not meet the AGV parameter requirements from the path graph, a breadth-first search is performed on the remaining channels to extract all valid reachable paths from the starting point to the target area to form a reachable path set. Each reachable path is represented by a node sequence, including the path number, path length, number of turns, and node coordinate set. Finally, a complete reachable path data set is formed to guide the start and end positioning of the transport task.

[0053] Step S164: Marking the starting position of the material based on the reachable path data to obtain the starting position data of the material; marking the target position based on the reachable path data to obtain the target position data;

[0054] In this embodiment, the starting position and the target position are extracted based on the spatial coordinate data bound to the cargo location. First, the cargo location number bound to the target cargo is retrieved in the three-dimensional model, and the three-dimensional coordinate center point (X, Y, Z) of the cargo location is extracted from the cargo location layout table. The coordinate point is then mapped to the voxel ID in the path grid, and the path node connected to the location is found through the transport channel map, marked as the "starting position node". The destination cargo location number is read from the task scheduling table, and the same operation is performed to extract the target location node. All node markings are performed in the path map, identified by the fields "is_start_node" and "is_target_node". The starting position data and the target position data are uniformly stored in JSON format. The fields include: cargo ID, cargo location number, voxel index, mapping channel node number, three-dimensional coordinates, and path number.

[0055] Step S165: Planning a transport path based on the starting position data and the target position data of the materials to obtain transport path data;

[0056] In this embodiment, the transport path planning uses the starting node and target node in S164 as endpoints and performs a heuristic path search in the reachable path graph. The path cost function is: F(n) = G(n) + H(n); where G(n) is the cumulative cost from the starting point to the current node n, and H(n) is the estimated cost from the current node to the target node (calculated using the Manhattan distance). The path cost coefficient is set as follows: the basic forward cost is 1.0, the 90° turn cost is 1.5, and a penalty weight of 0.5 is added when the channel width of each path segment is less than 1.2m. During the planning process, the total length of each path is controlled to be no more than 150m, and the number of turns is no more than 15. During the calculation process, if the path does not meet the AGV passability standards (such as the corner is too small), it is directly eliminated. After planning is completed, the transport path is output as a node sequence, including node number, coordinate value, cumulative cost, path direction sequence, etc. The transport path data is finally output in CSV format, with each line representing a node on the path, accompanied by the path number and task number index.

[0057] Step S166: Perform material transport simulation based on the transport path data to obtain material transport data.

[0058] In this embodiment, the transport simulation uses the path node sequence as the motion trajectory input, and models the AGV as a rigid body physical object. The control parameters include: starting acceleration of 0.3 m / s 2 , the uniform speed is 1.2m / s, and the maximum braking deceleration is 0.4m / s 2, the turning rate is limited to 0.8m / s. Each path segment is divided into three stages: acceleration, constant speed, and deceleration for speed timing distribution. The handling simulation execution process updates the status at a time interval of Δt=0.5s, and each update records the AGV's current position (X, Y, Z), speed, acceleration, direction angle, load status and other data. If an obstacle avoidance node is set in the path, a waiting time is inserted during the simulation process, and the default waiting time is 5.0s. After the simulation of the entire handling path is completed, the complete trajectory data is exported, including: [timestamp, AGV number, task ID, current position, speed, status identification]. The final generated material handling data is used to support subsequent collision prediction, path congestion assessment and system scheduling optimization. The storage format is binary log and JSON index dual format synchronization, which is convenient for data compression and parsing.

[0059] Preferably, the cargo placement analysis in step S2 is specifically as follows:

[0060] Calculate cargo flow density based on high-frequency congestion areas;

[0061] In this embodiment, equilateral grid units are constructed in the internal area of ​​the warehouse, and the side length of each grid is fixed at 3 meters. By deploying fixed ultra-high frequency RFID reading devices at key nodes (such as entrances and exits, and main channel intersections), identification data is collected once a second, and the timestamp, reading location number, and cargo ID are recorded. The number of reads in each grid unit for 20 consecutive minutes is counted, and the total number of reads is used to represent the cargo circulation intensity of the grid per unit time. When the number of reads of a grid exceeds 90 times within this time period and is maintained for more than 3 consecutive cycles, the area is marked as a high-frequency congestion area. The data is recorded through the warehouse digital control platform and is structured and stored with time segments as indexes. The fields include grid number, reading frequency, corresponding time period, cargo ID aggregation record, and reading device number.

[0062] Identify key transportation channels based on cargo circulation density and obtain key transportation channel data;

[0063] In this embodiment, all high-frequency congestion area grids in the previous step are analyzed for continuity according to the spatial adjacency rule, and the four-directional adjacency rule is used to connect the dense areas to extract the continuous linear direction areas. Paths with more than 5 consecutive grids and a length of more than 6 meters are defined as critical transportation channels. The identification of critical transportation channels uses a spatial adjacency clustering processing tool to mark each channel with its starting point, end point, grid number, and connection relationship, and construct a structured data set for archiving and management. The key transportation channel data is used as an input field in the path planning submodule to participate in the subsequent path priority analysis. The fields include channel number, start and end coordinate points, channel length, and area number.

[0064] Calculate cargo flow frequency based on key transportation channel data;

[0065] In this embodiment, within the key transport channel, each channel is counted for the number of times goods pass through it in 10-minute time units. Each pass is confirmed by the sequential reading time and ID of two fixed readers in front and behind. Only when the reading interval is less than 120 seconds and the direction is consistent, it is counted as a valid pass. The number of valid passes of each channel within 12 consecutive time units (i.e., 2 hours) is totaled as the frequency record of the goods corresponding to the channel. The frequency statistics field includes the channel number, statistical time segment, cargo ID, number of passes, direction number, timestamp sequence, etc. The data is written to the analysis buffer area and bound to the subsequent cargo tag data structure mapping.

[0066] Identify high-frequency cargo identification according to the frequency of cargo traffic and obtain high-frequency cargo identification data;

[0067] In this embodiment, the cumulative number of times all goods pass through key transportation channels is summarized, aggregated, and sorted by cargo ID. If a particular cargo passes through a key transportation channel more than 30 times within a 24-hour period, or more than 10 times within any three-hour period, the cargo ID is marked as a high-frequency cargo. This identification rule is embedded in the server's circulation identification rule engine and automatically executes batch marking at 2:00 AM daily. High-frequency cargo identification data includes the cargo's unique ID, label EPC code, route number, cumulative number of passes, trigger time, and location. All marked cargo is automatically assigned to a high-frequency cargo zone list.

[0068] Cargo priority is divided according to high-frequency cargo identification data to obtain cargo priority data;

[0069] In this embodiment, the system assigns a weighted score based on the frequency of goods' movement and time in the warehouse. Specifically, goods with frequent entry and exit and short storage time are designated as "high priority" goods; goods with moderate or intermittent entry and exit frequency are designated as "medium priority"; and the rest are classified as "low priority." The scoring criteria are as follows: goods with 30 or more passes in 24 hours and an average single storage time of less than 2 hours are designated as "high priority"; goods with 10 to 29 passes are designated as "medium priority"; and goods with fewer than 10 passes are designated as "low priority." The classification results are stored in the goods attribute database, with corresponding fields including the goods ID, priority level, score, latest entry and exit time record, and current storage location number.

[0070] Calibrate adjacent shelves based on cargo priority data to obtain adjacent shelf data;

[0071] In this embodiment, based on the priority results, shelf units near the main aisles and within 12 meters of the entrance and exit centers are screened from the intelligent warehouse structure diagram and designated as "priority placement areas." The available shelf space is then scanned. Each shelf's proximity is calculated using spatial geometric distance. Shelves are designated as "proximate" when their distance from the centerline of a high-priority aisle is less than 3.5 meters and their shelf level availability is greater than 60%. All adjacent shelf numbers are entered into a storage location allocation list and dynamically bound to the corresponding priority cargo identifiers.

[0072] The cargo placement simulation is performed based on the data of the adjacent shelves, where the handling safety margin is set to ≥0.6m and the placement direction is set to ≤15° to obtain the cargo placement data.

[0073] In this embodiment, a simulated handling task configuration is performed in an area adjacent to the shelf, and the geometric volume parameters and starting position data of the goods are called in using the three-dimensional warehouse simulation console. The minimum margin between each piece of goods during the placement process is set to be no less than 0.6 meters as a safe handling distance to prevent path obstruction and turning interference. The deviation angle of the placement direction of the goods is controlled within 15 degrees, that is, the maximum allowable angle relative to the shelf edge does not exceed 15 degrees. The scheduling system automatically calculates the position coordinates, rotation angle and placement level of the goods on the shelf based on the above two parameter constraints. The placement data fields output by the simulation platform include: goods ID, shelf number, X / Y / Z coordinates, rotation angle, stacking order, safety margin value, and timestamp.

[0074] Preferably, the identification of the cargo space form in step S2 is specifically as follows:

[0075] Identify cargo boundaries based on cargo placement data to obtain cargo boundary data;

[0076] In this embodiment, the XYZ coordinates, dimensions (length, width, and height), rotation angle, etc. of each product in the goods placement data are used as input and loaded into the warehouse visualization boundary recognition system. A rasterization space segmentation method is used to divide the three-dimensional space into units of 0.01m, and a minimum enclosing cube is established at the center of each cargo voxel. Collision detection is performed on the intersection area between the enclosing box and the shelf boundary. The actual boundary point positions are extracted based on the six-sided boundary point set of the cargo, and a six-sided boundary point coordinate matrix is ​​generated. If there is an area with a discrete point distance greater than 0.1m on the same coordinate plane, edge filling processing is required, and the nearest neighbor expansion strategy is used to fill in the boundary information to ensure that the boundary is completely closed.

[0077] Cargo voxel segmentation is performed based on cargo boundary data to obtain cargo voxel data;

[0078] In this example, the cargo boundary point cloud data output from the previous stage is imported into the voxelization processing engine and partitioned into three-dimensional blocks at a granularity of 0.01m×0.01m×0.01m. Each voxel unit is marked as either "occupied" or "vacant." During the voxel filling process, a three-dimensional flood fill scanning algorithm is used. Starting from the center of the bounding box, the marked voxels are filled in from the inside out, with only those within the boundary being marked as occupied. After segmentation, attribute information such as the cargo number, spatial coordinates, and shelf number are attached to each voxel. A structured voxel list is then output, with fields including the voxel ID, 3D coordinates, cargo ID, and boundary adjacency status.

[0079] Reconstruct the outer contour of the cargo according to the cargo voxel data to obtain cargo contour data;

[0080] In this example, for each calibrated set of cargo voxels, mesh data is generated using an eight-neighborhood voxel connectivity graph. A voxel boundary topology map is established at a 0.01 μm granularity. An isometric extraction operation is performed on the voxel exterior surface, connecting the common boundary points of all adjacent surface voxels to form a triangular mesh outline. The Marching Cubes topology reconstruction algorithm is used to perform continuity checks and contour closure on all meshes, ultimately outputting a mesh set with normal vector directions. Each piece of cargo outline data contains structural fields such as a set of boundary coordinate points, a normal vector direction, a vertex sequence, a closure status bit, and the corresponding cargo number.

[0081] Cargo placement symmetry analysis is performed based on cargo outline data to obtain cargo placement symmetry data;

[0082] In this embodiment, a spatial point set is constructed using the center points of all facets in the contour data, and mirror comparison is performed in the X, Y, and Z axis directions. During the mirror comparison process, the offset error after the center point is symmetrically projected about a certain axis is calculated. When the mean error in a certain direction is less than 0.03m and the maximum error does not exceed 0.05m, the direction is judged to be symmetrical. Three directional error threshold standards are set in the analysis: 0.03m for the X axis, 0.03m for the Y axis, and 0.04m for the Z axis. The result is marked with a Boolean flag to indicate the three-axis symmetry status. The record fields include the cargo ID, X / Y / Z axis symmetry flag, maximum offset value, mean error, number of corresponding symmetrical projection point pairs, etc.

[0083] Identify the cargo center axis based on the cargo placement symmetry data and obtain cargo center axis data;

[0084] In this example, a point set of biaxially symmetrical cargo voxels is extracted from the symmetry analysis results. The geometric center of the principal symmetry direction is calculated based on the weighted density of the point set. A least-squares fitting method is used to extract the axis of symmetry along the identified direction, and the coordinates of the starting and ending points of the principal axis, as well as the direction vector, are calculated. The central axis extraction process must meet the following conditions: the fitting residual must not exceed 0.02m, and the continuous length must exceed 80% of the maximum cargo dimension. The extraction results contain a structure containing fields such as the cargo number, the coordinates of the starting and ending points of the central axis, the direction vector, the fitting error value, the axis length, and the cargo location number where the axis is located.

[0085] Determine the main extension direction of the cargo based on the cargo centerline data to obtain the main extension direction data;

[0086] In this embodiment, the angle between the direction vector of the cargo's centerline and the three reference vectors of the warehouse's X, Y, and Z directions is compared, and the direction with the smallest angle is selected as the cargo's main extension direction. When the angle in a certain direction is less than 30° and the ratio of the projected length of the main axis in that direction to the total length is greater than 80%, that direction is marked as the main extension direction. The data output fields include the cargo ID, the main extension direction code (X / Y / Z), the direction angle value, the axis projection length, and the comparison direction angle. The extension direction results are used for subsequent three-dimensional structure alignment and cargo location sorting.

[0087] Perform cargo 3D structure registration based on the main extension direction data to obtain cargo 3D structure data;

[0088] In this embodiment, the current coordinate system of the goods is rotated based on the main extension direction to establish a local coordinate system, aligning the extension direction with the positive X-axis and the top surface facing the positive Z-axis. During the posture adjustment process, the coordinates of all contour vertices are reconstructed using the rotation matrix mapping to ensure that all goods maintain a consistent orientation. Euler angle rotation is used during this process, with each axis rotated no more than 90 degrees with an accuracy of 0.1 degrees. After the rotation is completed, the patch boundaries are regenerated. The output 3D structure data fields include the coordinate point set after the uniform orientation, the surface normal vector, the boundary face number, and the goods dimension bounding box.

[0089] The cargo spatial form is restored according to the cargo three-dimensional structure data to obtain cargo spatial form data.

[0090] In this embodiment, the three-dimensional structure after registration is topologically filled, and the boundary surface closure detection algorithm is used to check the consistency of all boundary areas to ensure that the contour is closed and complete. The registered coordinate data is imported into the three-dimensional structure assembly system, and a complete three-dimensional volume is established for the structure in the form of a monomer bounding box. The spatial morphological data is described using a multi-layer structure, and the levels include: basic information layer of goods (ID, cargo location number), geometric morphological layer (length, width, height, volume), structural layer (surface normal vector, boundary curvature) and space occupancy layer (voxel density distribution). Finally, the morphological data is input into the visualization platform for panoramic restoration of the goods.

[0091] Preferably, the cargo gravity center offset detection in step S2 is specifically as follows:

[0092] Identify the cargo bottom structure based on cargo spatial morphology data to obtain cargo bottom structure data;

[0093] In this embodiment, after obtaining the complete three-dimensional spatial morphological data of the cargo, the Z-axis downward projection method is used to extract the plane where the lowest point is located. Traverse the three-dimensional coordinate point set and select all points whose Z values ​​are within the range of ±0.005m of the minimum Z coordinate of the cargo as the bottom surface candidate point set. After performing a two-dimensional projection on the point set, use the convex hull extraction algorithm (such as Graham scan) to extract its minimum enclosing boundary and construct a two-dimensional bottom surface polygonal outline. The bottom surface structure data must include a set of boundary points, a plane normal vector (standard is 0,0,–1), and a bottom surface area (unit: m 2 , the error does not exceed ±0.01m 2 ), the number of boundary points, the distribution of boundary curvature, and other information. This operation requires the use of a spatial point cloud processing module, with a calculation resolution controlled at the millimeter level.

[0094] Perform contact surface shape recognition based on cargo bottom surface structure data to obtain contact surface shape data;

[0095] In this embodiment, based on the bottom surface structure data, the bottom surface projection plane is divided into 0.01m×0.01m grid units, and the standard deviation of the Z value of each point in the grid is calculated. If the standard deviation is ≤0.002m, the area is determined to be a flat surface. The adjacent flat grids are merged using the region growing method to extract the overall contact surface area and record the contact surface shape features. If the formed contact surface boundary is a regular rectangle or an approximate rectangle (side length error ≤0.02m, angle error ≤3°), the shape is identified as a "rectangular plane", otherwise it is identified as an "irregular surface" or "polygonal surface". The output contact surface shape data should include fields such as shape type, boundary vertex coordinates, side length, angle, area, and convexity index (calculated by the change in boundary curvature).

[0096] evaluating stacking adaptability according to contact surface shape data to obtain stacking adaptability data;

[0097] In this embodiment, the contact surface shape data is compared with the shelf support surface structure. The support surface boundary data of the target shelf platform or the top of the upper cargo is read, and the boundary projections of the two are aligned in the same coordinate system. The boundary matching rate calculation method is adopted: the ratio of the overlapping area of ​​the cargo contact surface and the load-bearing surface to the total contact area is set as the matching rate. If the matching rate is ≥90% and the center of gravity of the contact surface is within the boundary of the load-bearing surface, the stacking adaptability is "adaptive"; if the matching rate is between 70% and 90%, it is marked as "low adaptability", and the rest are "unfitting". The output stacking adaptation data must include detailed fields such as adaptation level, overlapping area, deviation between the contact center point and the load-bearing surface center point, and contact boundary coverage.

[0098] Determine the cargo support gravity center based on the stacking adaptation data to obtain the support gravity center data;

[0099] In this embodiment, the centroid points of all flat areas are extracted from the projection plane of the contact surface, and the overall support center of gravity position is calculated according to the area and shape distribution of each area using the area weighted method. Each contact area is regarded as an equal mass distribution body, and its center of gravity position is taken as the representative point, and the calculation method is weighted coordinate averaging. The Z coordinate of the support center of gravity is uniformly set to the bottom height of the cargo (i.e., the minimum Z value). If the contact surface is an irregular polygon, the area segmentation method is used to divide it into multiple triangular units, and the center of gravity of each triangle is calculated separately and then merged and weighted. The support center of gravity data should record the center of gravity coordinates (X, Y, Z), the center of gravity offset, the horizontal distance from the projection point of the center axis of the cargo, and a Boolean flag of whether it is within the load-bearing surface.

[0100] Calculate the spatial offset vector based on the support center of gravity data;

[0101] In this embodiment, a spatial offset vector is constructed with the center of mass of the cargo (obtained by weighting the complete spatial voxel point set) as the starting point and the center of gravity of the support as the end point. The components of the vector in the X, Y, and Z directions are calculated, and the horizontal projection length (√(ΔX2+ΔY2)) and the vertical offset ΔZ value are specially marked. The offset vector also needs to calculate the angle with the direction of gravity (negative direction of the Z axis), and record the offset direction as "front / back / left / right" or a combination thereof. The output vector data field includes the offset start point coordinates, end point coordinates, three-dimensional components, unit vectors, horizontal offsets, vertical offsets, and angles (accuracy controlled within 0.1°).

[0102] The cargo center of gravity offset is detected based on the spatial offset vector, where a cargo tilt angle ≥ 5° is judged as a skew risk, and the cargo center of gravity offset data is obtained.

[0103] In this embodiment, the calculated offset vector data is read, and the ratio of the length of its horizontal component to the overall height of the goods is calculated to further obtain the tilt angle θ=arctan(horizontal offset / height). If θ≥5°, ​​"there is a risk of deflection" is marked in the detection data. The detection process uses the standard critical angle threshold set to 5°, and the judgment basis is the stability reference limit of the national standard GB / T4857.10-2005 "Test method for transport of packaged goods". The output center of gravity offset data must include: tilt angle, offset direction, offset vector length, difference between the tilt angle and the standard critical angle, whether the risk threshold Boolean value is reached, offset risk level (low / medium / high), and the corresponding cargo number and stacking layer number.

[0104] Preferably, step S3 is specifically as follows:

[0105] Step S31: Collect the unique label of the cargo according to the cargo gravity center offset data to obtain cargo label data;

[0106] In this embodiment, an ultra-high frequency RFID reader / writer is used to extract data from the embedded RFID electronic tag on the surface of the outer packaging of the goods. The tag contains a unique identifier in the form of a 96-bit EPC code, and the coding structure includes an 8-bit manufacturer prefix, a 20-bit product type, a 44-bit serial number, and a 24-bit anti-counterfeiting verification segment. During the extraction process, the reader antenna power is set to 30dBm, and the reading distance is set to 2.5 meters. The reading process requires multiple verifications to ensure that the tag information is complete and correct. After the reading is completed, the tag information needs to be bound to the current cargo center of gravity offset coordinate data (including X, Y, Z three-dimensional direction offset, inclination information, etc.) to generate a complete cargo tag data structure and write it into the local intermediate database for upload.

[0107] Step S32: Upload the cargo label data to the warehouse management system, and display the cargo storage status to obtain a real-time storage status diagram;

[0108] In this embodiment, the upload process uses the MQTT protocol to publish information at QoS2 level within the local area network to ensure that the message transmission is complete and without duplication. The uploaded content includes fields such as tag ID, collection timestamp, bound goods ID, location information, and center of gravity offset data. After receiving the data, the system uses the WebSocket protocol to push the data to the digital twin visualization terminal and displays it in a stateful manner in a three-dimensional digital warehousing environment. The display process is based on the Unity or Cesium platform. In the rendering process, the point cloud layer is constructed to overlay the label node, and the state diagram node is constructed according to the location information recorded in the label. The real-time warehouse status diagram is rendered in combination with different color identifiers (such as green for stability, orange for offset risk, and red for overload risk).

[0109] Step S33: identifying the type of goods based on the real-time warehouse status diagram to obtain goods type data; performing a reflective inspection of the goods material based on the goods type data to obtain goods material reflective data; and reconstructing a depth map based on the goods material reflective data to obtain a goods depth map;

[0110] In this embodiment, the cargo category database is associated with the tag ID, and the category information matching the cargo number is read therefrom. If it is identified as a reflective package (such as a metal film composite bag or a PE coated bag), a material reflective detection is required. A line laser projection device is used to scan the cargo, and a reflectivity sensor is used to obtain the surface reflection intensity value. The reflectivity threshold is set to 0.75. When the area detected above the threshold exceeds 30% of the surface area of ​​the cargo, it is marked as a "strong reflective package". Subsequently, based on this reflective information, a dual-camera stereo vision system is used to collect reflective correction images at different angles, the parallax method is used to invert the image depth, and the triangulation method is used to reconstruct the depth map, and finally the cargo depth map data is generated.

[0111] Step S34: extracting standard weight information of the goods based on the goods type data; extracting a goods image based on the real-time warehouse status diagram; identifying wrinkles in the goods packaging based on the goods image to obtain wrinkle data; and correcting the goods volume based on the wrinkle data to obtain corrected goods volume data;

[0112] In this embodiment, the standard weight information of the corresponding goods is extracted from the standard product database using the cargo type information. The data sources include historical warehouse records and information provided by the manufacturer. The unit is kilogram and the accuracy is 0.01kg. At the same time, the image data of the corresponding goods is extracted from the warehouse status map (based on the RGB image stream, 6 frames per second). The edge gradient-based image recognition algorithm is used to identify wrinkles on the packaging surface. The wrinkle recognition conditions are that the local gradient change exceeds ±40 grayscale units and the area is larger than 5cm. 2 After wrinkle areas are identified, the system uses the original packaging volume data (via barcode or system settings) to perform volume corrections. The volume correction formula relies on wrinkle depth and area assessment, with the correction ranging from 2% to 8% of the original volume. The specific value is estimated based on the cumulative effect of wrinkle depth and affected area, and the final output is the corrected volume data for the goods.

[0113] It is particularly important that step S34 includes the following steps:

[0114] Step S341: extracting cargo standard weight information based on cargo type data;

[0115] In this embodiment, the cargo identification module is called in the IoT terminal built into the warehousing system, and the cargo type code is read through the RFID tag, barcode or QR code. According to the code, the corresponding cargo standard attribute table is queried in the warehousing database, which contains the standard weight information corresponding to each cargo type. The standard weight information is stored in "kg" with an accuracy of 0.01kg. For goods with different packaging forms, the standard weight information is subdivided and searched according to "cargo type + packaging specifications" to ensure that the search results are unique. All standard weight data are filed by the manufacturer when entering the warehouse and are manually reviewed and confirmed. The final output is the standard weight field, such as: {Cargo ID: GZ202506, standard weight: 17.85kg}.

[0116] Step S342: extracting cargo images based on the real-time warehouse status map;

[0117] In this embodiment, image data of the corresponding cargo area in the real-time warehouse status diagram is extracted in the visualization module of the warehouse management system, and image acquisition is performed using a 5-megapixel industrial camera fixedly mounted on the shelf beam. The size of each image frame is 2592×1944 pixels, and the acquisition frequency is 1fps. The image acquisition channel uses the industrial Ethernet protocol to connect to the main control platform of the warehouse system to ensure that the delay does not exceed 50ms. For each cargo image area, the image segmentation algorithm is called to perform background removal based on a predefined pixel label mask, retaining the complete image of the cargo outline as input for subsequent processing. Each image needs to save information such as the image number, timestamp, and corresponding cargo label ID for subsequent data synchronization processing.

[0118] Step S343: Calculating texture gradients based on the cargo image to obtain texture gradient data; identifying texture mutation boundaries based on the texture gradient data to obtain suspected wrinkle boundary data;

[0119] In this embodiment, the cargo image data obtained in step S342 is used to call a texture gradient extraction algorithm based on the combined calculation of the gray-level co-occurrence matrix (GLCM) and the Sobel operator. First, the image is grayscaled and the gradient changes of the image on the X and Y axes are calculated using a Sobel operator with a 3×3 kernel to generate a two-dimensional texture gradient map. Secondly, a local sliding window with a window size of 21×21 pixels is used to analyze the local texture difference and calculate the energy, contrast, and homogeneity in the gray-level co-occurrence matrix. The texture mutation threshold is set to a gradient change rate greater than 35 (unit pixel grayscale / pixel distance), and the boundary position of continuous texture changes is located in the image. Finally, the area that meets the mutation conditions is marked as the suspected wrinkle boundary area, and the output is the suspected wrinkle boundary coordinate set {(x1, y1), (x2, y2), ...}.

[0120] Step S344: calculating the spatial concavity based on the suspected fold boundary data; determining the fold boundary of the suspected fold boundary data based on the spatial concavity to obtain fold data;

[0121] In this embodiment, based on the suspected wrinkle boundary data identified in step S343, the binocular vision system is used to reconstruct the depth of the corresponding image area. The baseline distance of the binocular camera is set to 10 cm, the focal length is 4.2 mm, and the disparity map of the boundary area is extracted by the stereo matching algorithm (Semi-Global Matching, SGM), and then the Z-direction depression depth of the boundary point is calculated. The depression depth is calculated by the formula D = (f × B) / d, where f is the focal length, B is the baseline length, and d is the disparity pixel value. If the depression depth is greater than 3 mm and the continuous area is greater than 20 cm 2 , the boundary is determined to be a true wrinkle boundary. Valid boundaries are screened based on this criterion, and the wrinkle morphology is drawn in the 3D point cloud based on the depth map. The screened wrinkle area is ultimately output as a 3D wrinkle data set, including spatial coordinate range, maximum concave depth, and area index.

[0122] Step S345: Correct the cargo volume based on the wrinkle data to obtain corrected cargo volume data.

[0123] In this embodiment, the three-dimensional wrinkle area obtained in step S344 is spatially overlapped with the original volume model of the cargo. The original volume of the cargo is obtained through laser ranging or structured light scanning data. The original volume data is in three-dimensional point cloud or grid format with a resolution of 1mm. Based on the point cloud Boolean difference operation, the volume depression value corresponding to the wrinkle area is calculated. The discrete voxel grid (Voxel Grid) method is used for each wrinkle area, and the voxel side length is set to 5mm. The number of depressed voxels is counted and converted into volume values. If the cargo packaging material is a compressible material (such as plastic film, soft box, etc.), the packaging material correction coefficient K is introduced. This coefficient is set in advance according to the compression rate experiment, for example, K = 1.12 for PVC film and K = 1.05 for corrugated box. Finally, the formula Vcorrected = Voriginal-K×Vwrinkled is used to correct the cargo volume and output the cargo corrected volume data in cm. 3 , keep two decimal places.

[0124] Step S35: Calculate the actual stacking volume of the cargo based on the cargo depth map and the cargo corrected volume data; calculate the actual stacking weight of the cargo based on the actual stacking volume and the cargo standard weight information;

[0125] In this embodiment, the depth map is converted into point cloud data, and the point cloud density is controlled to be more than 100,000 points per cubic meter. The point cloud data is partitioned into voxels using an octree structure to remove background noise points and non-cargo area points. After voxel aggregation, the bounding box volume is calculated, and the volume correction data is combined to compensate for the smaller or larger values. The final output is the actual stacking volume of the cargo (unit: m 3 , with an accuracy of 0.001m 3 This volume value is then multiplied and converted with the standard weight information, with the density range set within a ±15% tolerance to ensure rationality. The actual stacking weight of the goods is ultimately calculated in kg with an accuracy of 0.01kg.

[0126] Step S36: Acquire shelf load limit data; perform shelf overload anomaly analysis based on the shelf load limit data and the actual stacking weight of the goods to obtain shelf overload anomaly data.

[0127] In this embodiment, the shelf load limit data is derived from the shelf design parameters, and the unit is kg / m 2 , establish a query table with the shelf level number as the index field. Read the shelf number through the system interface and extract the standard load-bearing value of the corresponding level. According to the actual stacking weight data of the goods, divide it by the shelf area occupied by the goods to obtain the unit area weight. Compare the unit area weight with the shelf load-bearing limit. If it exceeds 90% of the upper limit, it is marked as "near overload". If it exceeds the upper limit, it is recorded as "overload abnormality" and the cargo node is marked with a red flashing signal in the status diagram. At the same time, an abnormal data log is generated. The fields include cargo ID, stacking weight, shelf number, load-bearing limit value, detection time, etc. The log is stored in the system background database for subsequent query and disposal.

[0128] Preferably, step S4 is specifically as follows:

[0129] Step S41: performing fatigue detection on shelf connection nodes based on shelf overload abnormality data to obtain connection node fatigue data;

[0130] In this embodiment, the stress state of the nodes is monitored in real time by deploying multi-channel strain sensors on the key connection parts of the shelf structure (such as the junction of the main and secondary columns, and the contact surface between the beam and the column). A triaxial strain gauge is deployed at each node with a gauge spacing of 0.5 mm and a sampling frequency of 200 Hz. The data is collected using a wireless strain collector and uploaded to the edge computing gateway. The node strain mean and the maximum periodic fluctuation amplitude are used as fatigue assessment indicators in the analysis. The fatigue detection threshold is defined as ±600 με. If the value is exceeded continuously for 10^6 times, it is determined to be a critical fatigue state. The shelf number and detection time are recorded as a fatigue detection log, and the fatigue data of the connection node is output. The data fields include node number, axial strain, normal strain, force frequency, cumulative fatigue times and fatigue level.

[0131] Step S42: performing structural deformation detection based on the shelf overload abnormality data to obtain structural deformation data;

[0132] In this embodiment, a structured light 3D scanning device (resolution 0.2mm, working distance 2m) is used to scan the shelves in abnormal areas, and the structural information of the front, side and back of the shelf is collected by horizontal movement of the scanning head. The scanning data is spliced ​​through point clouds to obtain the complete three-dimensional shape of the shelf. Then, the CAD design standard shelf structure drawing is called and compared with the current scanned shelf point cloud, and the offset values ​​of the key structural points (such as the top of the center column and the center point of the beam) in the X, Y, and Z directions are calculated through the registration algorithm. The structural offset threshold is set to ±4mm, and when the offset exceeds the threshold, it is marked as a structural deformation area. A structural deformation data table is generated for each detection. The fields include part number, offset value, offset direction, offset area and deformation level, and the data is uploaded to the Internet of Things warehousing cloud platform.

[0133] It is particularly important that step S42 includes the following steps:

[0134] Step S421: collecting shelf structure images based on shelf overload abnormality data;

[0135] In this embodiment, upon receiving abnormal shelf overload data, the physical area corresponding to the abnormal data must be precisely located. Industrial visual acquisition equipment deployed in the warehouse environment is then used to capture images of this localized structural area. The acquisition equipment must be equipped with an industrial-grade CMOS image sensor with a resolution of at least 2448×2048 pixels and a sampling rate of 30 frames per second to ensure high-precision capture of subtle deformations. The image acquisition range is limited to the shelf number and level index indicated in the overload data, and the target area is precisely located using a pre-set warehouse coordinate index mapping table. The image acquisition process is performed using a standard LED array light source, maintaining an illumination intensity between 35,000 and 40,000 lux to avoid image loss caused by reflective shelf materials or localized shadows. The captured images must be saved in uncompressed .tiff format and immediately synchronized to the edge data processing node for the next round of boundary recognition. The image acquisition window is limited to 5 seconds after an abnormality detection occurs to ensure the continuity and timeliness of the detection and feedback process.

[0136] Step S422: identifying the structural contour boundary based on the shelf structure image to obtain structural contour boundary data;

[0137] In this embodiment, after image acquisition is completed, edge enhancement processing is performed using the grayscale features of the structural image, specifically using the Canny algorithm to implement a multi-threshold edge extraction operation. The dual thresholds are set to low threshold = 80, high threshold = 200, and the image convolution kernel size is set to 3×3. Based on the extraction results, the Hough transform method is used to identify regularly arranged straight feature line segments in the image. Such line segments correspond to the main structural units of the shelf contour boundary. In this process, the boundary line is corrected by coordinate projection mapping to eliminate the perspective distortion caused by the deviation of the shooting angle. The boundary point extraction needs to control the pixel error to no more than ±2 pixels. After the extraction is completed, the coordinates of all structural boundary points are uniformly stored in a two-dimensional boundary data set. The boundary data set format is defined as {boundary point ID, X coordinate, Y coordinate, extraction confidence}. Boundary points with a confidence level less than 0.85 will be judged as non-structural boundaries and eliminated to ensure the stability and accuracy of boundary recognition.

[0138] Step S423: Calculating boundary node spatial coordinates based on the structure outline boundary data;

[0139] In this embodiment, after obtaining the structural contour boundary data, the image coordinates are back-projected in three dimensions using the preset camera internal parameter matrix and external parameter matrix to convert them into the spatial coordinates of the boundary nodes. The internal parameters include focal length f = 16 mm, principal point coordinates (cu, cv) = (1224, 1024), and distortion parameters are obtained through camera calibration; the external parameter matrix contains rotation vectors and translation vectors, whose values ​​are calibrated during deployment and saved to the local configuration file. The three-dimensional coordinate back-projection uses a pinhole camera model combined with image depth information for point-by-point solution. The depth information is synchronously acquired from an industrial-grade laser ranging sensor, and the measurement accuracy is required to be ±0.3 mm. The converted spatial coordinates of each boundary point are expressed as {X, Y, Z}, and the results are uniformly stored as a three-dimensional node dataset of the structural boundary, and node identification labels are added for subsequent difference comparison operations. The spatial coordinates of all nodes will be uniformly transferred to the structural analysis module to enter the spatial difference calculation process.

[0140] Step S424: Calculate the difference between the preset standard node space coordinates and the boundary node space coordinates to obtain node offset data;

[0141] In this embodiment, for the spatial coordinates of the boundary nodes obtained in step S423, the system's built-in shelf standard structure model coordinate library is called to perform point-to-point coordinate comparison. The standard model is a CAD structural three-dimensional benchmark model, and the number of nodes and naming rules completely correspond to the physical deployment of the shelf. The comparison operation uses Euclidean distance as the calculation basis, and the offset calculation formula for each pair of nodes is: Δd = √[(X1-X0)2+(Y1-Y0)2+(Z1-Z0)2], where (X0, Y0, Z0) are standard coordinates and (X1, Y1, Z1) are measured coordinates. In the node offset data, nodes with Δd exceeding 3.5mm are marked as "abnormal offset", and the offset value and node number form a structured data table. The threshold is set by the system based on mechanical testing and historical warehouse data experience. The sources of error include local plastic deformation and loose connection points. After this step is completed, the complete node offset data set will serve as the basis for calculating the local deformation in the next step.

[0142] Step S425: determining the local deformation amount based on the node offset data; reconstructing the deformed structure according to the local deformation amount to obtain structural deformation data.

[0143] In this embodiment, based on the node offset data, the structural unit composed of continuous nodes is selected to calculate the regional deformation variable. The offset node area is divided into patches using a double triangle mesh topology method, and each patch unit is represented as a three-node plane structure. The deformation variable ΔS uses the change in the patch center of gravity as the core indicator, and the formula is ΔS = |S1-S0|, where S0 is the standard structure area and S1 is the actual area after deformation. The change in the unit area of ​​the patch needs to be controlled within the analysis accuracy of ±0.05cm2 , as the quantitative analysis result of the structural unit deformation. Subsequently, all facets are globally fitted according to the structural symmetry and spatial connection topology to construct the three-dimensional geometry of the deformed body. The reconstruction process relies on the spatial B-spline interpolation reconstruction algorithm, which repairs missing points through local node interpolation calculation to form a complete deformed structural mesh. Finally, the deformed structural mesh is exported as a .stl format three-dimensional data file, and the local deformation values ​​of all facets are recorded to represent the final structural deformation data. This data is used as the input source for structural health analysis for subsequent prediction tasks.

[0144] Step S43: integrating the connection node fatigue data and the structural deformation data to obtain the shelf structure fatigue data;

[0145] In this embodiment, the integration process is carried out in the edge server, using a unified structure number as the primary key field, and quickly matching the connection node fatigue data with its corresponding structural deformation data through a hash index. The structural fatigue level merging index is defined as follows: if a node has both a critical fatigue state (such as fatigue level Ⅲ) and a structural deformation offset value exceeding 6mm, the merged judgment is that the structural fatigue level is level Ⅳ, reflecting a high-risk structural state. In this process, it is also necessary to record the associated cargo ID, overload duration (minutes), and node corresponding position coordinates, and construct a shelf structure fatigue dataset, which is stored in JSON format, including structure number, node fatigue level, corresponding structure offset value, affected area, and stress concentration trend information.

[0146] Step S44: predicting shelf life based on shelf structure fatigue data to obtain shelf life data;

[0147] In this embodiment, the life calculation is completed in the cloud server, and the indicators based on which the analysis is based include: the number of node fatigue cycles, the node strain amplitude, the structural offset rate (mm / day), and the current usage time (days). The number of node fatigue cycles comes from the accumulation of historical strain data, the offset rate is linearly fitted based on the results of the last three deformation tests, and the current usage time is calculated based on the installation record timestamp. The life prediction logic adopts the threshold attenuation method: the initial structural design life is set to 1800 days. When the node fatigue exceeds the critical value of 80% and the offset rate exceeds 0.3mm / day, 3 days of life are deducted for each unit day exceeded. The final output is the remaining service life data of the shelf, and the fields include structure number, remaining life days, and remaining life level (such as I ≥ 500 days, II 200 to 499 days, III < 200 days), and a life database index table is established for all structure numbers for real-time query.

[0148] Step S45: Generate a shelf health heat map based on the shelf life data.

[0149] In this embodiment, the heat map is constructed based on the warehouse three-dimensional modeling platform (such as that implemented based on Cesium or Three.js), and a color layer is superimposed on the original coordinate system of the shelf to reflect the life level. The remaining life value is mapped in different color intervals: green represents level I (sufficient life), yellow represents level II (needs to be monitored), and red represents level III (warning state). In the layer generation, the shelf unit (such as a 1m×1m area) is used as the smallest unit, and the color filling is accurately corresponding to the thermal grid by the node coordinates, and it is automatically refreshed every 30 seconds to ensure that the display is synchronized with the life data. The color threshold control parameters are configured in the system and set as: green for more than 500 days, yellow for 200 to 500 days, and red for less than 200 days. The heat map data format is GeoJSON, which contains fields such as structure number, coordinates, life level, last detection time, etc., and can be viewed and exported in real time through the warehouse digital twin platform.

[0150] Preferably, step S41 is specifically as follows:

[0151] Step S411: extracting the overload effect layer according to the shelf overload abnormality data to obtain the shelf overload data;

[0152] In this embodiment, the shelf layer number information corresponding to the cargo location information is extracted by reading the three-dimensional coordinate calibration database of the shelf. According to the XYZ three-axis coordinates, each 300mm in the Z-axis direction is set as a layer, and the total height of the shelf is set to 2400mm, with a total of 8 layers. The Z-axis height of the cargo whose center of gravity offset exceeds ±5° or whose stacking weight is greater than the single-layer maximum load limit of 200kg is matched with the number to calibrate the overloaded layer. The corresponding layer is marked as an overloaded layer, and the layer overload data is generated. The data fields include the layer number, the corresponding cargo ID, the cargo mass value (kg), the center of gravity offset angle (°), the stacking volume (m 3 ) and overload duration (min). The data will be stored in the warehouse monitoring database and used for subsequent mechanical analysis.

[0153] Step S412: detecting the force conduction path based on the layer overload data to obtain the force conduction path data;

[0154] In this embodiment, a complete force flow path monitoring network is constructed using strain measuring instruments and tensile and compressive strain gauges deployed between the shelf columns and beams. The strain gauges are arranged at the support points of each shelf, the interface between the main and secondary shelves, and the connection position of the reinforcement ribs, and the strain acquisition frequency is set to 500Hz. During the loading process of the shelf, the strain conduction direction is calculated by recording the strain value change trend and time series. The three-dimensional finite point method is used to analyze the node force transfer, and the force is progressively transferred from the support point to the column according to the force increment trend, and a force conduction path diagram is constructed. In the path diagram, each path segment with a stress change gradient exceeding ±60με / mm is marked as a main conduction path. Finally, the force conduction path data is generated, and the fields include the path start and end nodes, node number sequence, stress change gradient (με / mm), total path length (mm) and average strain value.

[0155] Step S413: identifying stress concentration areas based on the force conduction path data;

[0156] In this embodiment, the stress increment values ​​corresponding to all path nodes are extracted in the force path. The stress concentration identification threshold is set to ±800με. If the stress changes of three consecutive nodes exceed this value, the area is marked as a stress concentration segment. During the analysis process, the sliding window method (the window length is set to 3 nodes) is used to compare the local stress changes in the path, and the local maximum strain difference is calculated. When the difference exceeds 200με, it is further determined that there is a high gradient concentrated stress. The three coordinate values ​​of the starting position, midpoint and end point of the stress concentration segment are extracted and marked as a stress concentration area. Generate stress concentration area data, and the data fields include path number, concentrated area number, center coordinates, local stress gradient value, concentrated area length (mm) and force level mark (Ⅰ, Ⅱ, Ⅲ).

[0157] Step S414: locating high-risk laminate connection nodes based on stress concentration areas;

[0158] In this embodiment, by comparing the layout of the connection points between the layers and the columns in the shelf structure drawings, the specific node numbers through which the stress concentration section passes are located. In the structural design, each layer is arranged with 4 connection nodes in the direction of the beam (left-right symmetrical) and 2 to 3 in the longitudinal direction. The spatial position of the concentrated area is compared with the node coordinates in three dimensions. If the center of the node is less than 50mm from the center point of the concentrated area and the corresponding stress of the node is greater than 1000με, it is marked as a high-risk connection node. At the same time, the connection method (welding / plug-in), node number, position number and connection plate thickness (such as 2.5mm) are recorded to generate a high-risk connection node data list for subsequent desoldering detection.

[0159] Step S415: performing desoldering detection on high-risk layer connection nodes to obtain desoldering data;

[0160] In this embodiment, an ultrasonic guided wave detection instrument is used to perform non-contact scanning of the welding area, the transmission frequency is set to 40kHz, and the scanning speed is set to 2mm / s. During the detection process, the connection node is judged whether there is a discontinuous weld, cracks or solder voids and other desoldering phenomena based on the ultrasonic echo delay and amplitude change. For plug-in type nodes, a high-precision displacement sensor is used to detect whether the node undergoes relative displacement during the loading process. If a vertical displacement of more than 0.5mm is detected under 50N loading conditions, it is determined that the connection is loose or failed. The desoldering data includes the node number, detection method, echo loss value (dB), structural crack length (mm), displacement (mm) and judgment result (qualified / desoldered), and a node desoldering detection data table is formed.

[0161] Step S416: determining the fatigue degree of the connection node according to the desoldering data, and obtaining the connection node fatigue data.

[0162] In this embodiment, the fatigue assessment method adopts a dual-factor weighted judgment strategy of "historical fatigue strain value + desoldering characteristic parameter". The sum of the cumulative strain values ​​at the node and the length of the desoldering crack are used as input parameters, the crack length influence weight is set to 0.7, the strain accumulation value influence weight is set to 0.3, and the node fatigue index F is calculated. Fatigue index F = 0.7 × (crack length / 10mm) + 0.3 × (strain accumulation value / 2000με). If F ≥ 1.2, it is judged as severe fatigue; 0.8 ≤ F < 1.2 is judged as moderate fatigue; F < 0.8 is mild fatigue. Record the fatigue level of the connection node and generate the fatigue data of the connection node. The data fields include node number, fatigue index, welding integrity assessment value, historical strain number, weld thickness, weld crack length, and final fatigue level.

[0163] Preferably, this specification also provides an Internet of Things-based intelligent warehouse digital visualization management system, which is used to execute the above-mentioned Internet of Things-based intelligent warehouse digital visualization management method, and the Internet of Things-based intelligent warehouse digital visualization management system includes:

[0164] The material handling simulation module is used to obtain intelligent warehouse structure data; build a three-dimensional warehouse model based on the intelligent warehouse structure data; and perform material handling simulation based on the three-dimensional warehouse model to obtain material handling data;

[0165] The cargo center of gravity offset detection module is used to identify high-frequency congestion areas based on material handling data; perform cargo placement analysis based on the high-frequency congestion areas to obtain cargo placement data; identify cargo spatial form based on the cargo placement data to obtain cargo spatial form data; and perform cargo center of gravity offset detection based on the cargo spatial form data to obtain cargo center of gravity offset data.

[0166] The shelf overload anomaly analysis module is used to collect unique cargo labels based on cargo center of gravity offset data to obtain cargo label data; upload cargo label data to the warehouse management system and display the cargo storage status to obtain a real-time warehouse status diagram; perform shelf overload anomaly analysis based on the real-time warehouse status diagram to obtain shelf overload anomaly data;

[0167] The shelf life prediction module is used to perform shelf structure fatigue detection based on shelf overload abnormal data to obtain shelf structure fatigue data; predict shelf life based on shelf structure fatigue data to obtain shelf life data; and generate a shelf health heat map based on the shelf life data.

[0168] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0169] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A digital visualization management method for intelligent warehousing based on the Internet of Things, characterized in that: The following steps are involved: Step S1: Obtaining intelligent warehousing structure data; Build a 3D warehouse model based on intelligent warehouse structure data; Conduct material handling simulation based on the 3D warehouse model to obtain material handling data; Step S2: Identify high-frequency congestion areas based on the material handling data; perform cargo placement analysis based on the high-frequency congestion areas to obtain cargo placement data; identify cargo spatial form based on the cargo placement data to obtain cargo spatial form data; perform cargo center of gravity offset detection based on the cargo spatial form data to obtain cargo center of gravity offset data; Step S3: Collect unique cargo labels based on cargo center of gravity offset data to obtain cargo label data; upload cargo label data to the warehouse management system, and display cargo storage status to obtain a real-time warehouse status diagram; perform shelf overload anomaly analysis based on the real-time warehouse status diagram to obtain shelf overload anomaly data; Step S4: Perform shelf structure fatigue detection based on the shelf overload abnormality data to obtain shelf structure fatigue data; predict shelf service life based on the shelf structure fatigue data to obtain shelf life data; and generate a shelf health heat map based on the shelf life data.

2. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 1 is characterized in that: Step S1 is specifically as follows: Step S11: Acquire intelligent warehousing structure data; Step S12: Identify building components based on the intelligent storage structure data to obtain building component data; Step S13: Divide the space units based on the building component data to obtain storage space partition data; Step S14: Reconstructing the cargo space layout according to the storage space partition data to obtain cargo space layout data; Step S15: Modeling the structural hierarchical relationship based on the cargo location layout data to obtain warehouse structure diagram metadata; assembling a three-dimensional space model based on the warehouse structure diagram metadata to obtain a three-dimensional warehouse model; Step S16: Perform material handling simulation based on the warehouse three-dimensional model to obtain material handling data.

3. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 2 is characterized in that: Step S16 is specifically as follows: Step S161: performing storage path gridding processing based on the three-dimensional warehouse model to obtain path grid data; Step S162: Identify the material transport channel according to the path grid data to obtain transport channel data; Step S163: Performing an AGV reachability analysis based on the transport channel data to obtain reachable path data; Step S164: Marking the material starting position based on the reachable path data to obtain material starting position data; Mark the target location based on the reachable path data to obtain target location data; Step S165: Planning a transport path based on the starting position data and the target position data of the materials to obtain transport path data; Step S166: Perform material transport simulation based on the transport path data to obtain material transport data.

4. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 1 is characterized in that: The specific analysis of cargo placement in step S2 is as follows: Calculate cargo flow density based on high-frequency congestion areas; Identify key transportation channels based on cargo circulation density and obtain key transportation channel data; Calculate cargo flow frequency based on key transportation channel data; Identify high-frequency cargo identification according to the frequency of cargo traffic and obtain high-frequency cargo identification data; Cargo priority is divided according to high-frequency cargo identification data to obtain cargo priority data; Calibrate adjacent shelves based on cargo priority data to obtain adjacent shelf data; The cargo placement simulation is performed based on the data of the adjacent shelves, where the handling safety margin is set to ≥0.6m and the placement direction is set to ≤15° to obtain the cargo placement data.

5. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 1 is characterized in that: The identification of cargo space form in step S2 is specifically as follows: Identify cargo boundaries based on cargo placement data to obtain cargo boundary data; Cargo voxel segmentation is performed based on cargo boundary data to obtain cargo voxel data; Reconstruct the outer contour of the cargo according to the cargo voxel data to obtain cargo contour data; Cargo placement symmetry analysis is performed based on cargo outline data to obtain cargo placement symmetry data; Identify the cargo center axis based on the cargo placement symmetry data and obtain cargo center axis data; Determine the main extension direction of the cargo based on the cargo centerline data to obtain the main extension direction data; Perform cargo 3D structure registration based on the main extension direction data to obtain cargo 3D structure data; The cargo spatial form is restored according to the cargo three-dimensional structure data to obtain cargo spatial form data.

6. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 1 is characterized in that: The specific detection of cargo gravity center offset in step S2 is as follows: Identify the cargo bottom structure based on cargo spatial morphology data to obtain cargo bottom structure data; Perform contact surface shape recognition based on cargo bottom surface structure data to obtain contact surface shape data; evaluating stacking adaptability according to contact surface shape data to obtain stacking adaptability data; Determine the cargo support gravity center based on the stacking adaptation data to obtain the support gravity center data; Calculate the space offset vector based on the support center of gravity data; The cargo center of gravity offset is detected based on the spatial offset vector, where a cargo tilt angle ≥ 5° is judged as a skew risk, and the cargo center of gravity offset data is obtained.

7. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 1 is characterized in that: Step S3 is specifically as follows: Step S31: Collect the unique label of the cargo according to the cargo gravity center offset data to obtain cargo label data; Step S32: Upload the cargo label data to the warehouse management system, and display the cargo storage status to obtain a real-time storage status diagram; Step S33: identifying the type of goods based on the real-time warehouse status diagram to obtain goods type data; performing a reflective inspection of the goods material based on the goods type data to obtain goods material reflective data; and reconstructing a depth map based on the goods material reflective data to obtain a goods depth map; Step S34: extracting standard weight information of the goods based on the goods type data; extracting a goods image based on the real-time warehouse status diagram; identifying wrinkles in the goods packaging based on the goods image to obtain wrinkle data; and correcting the goods volume based on the wrinkle data to obtain corrected goods volume data; Step S35: Calculating the actual stacking volume of the cargo based on the cargo depth map and the cargo corrected volume data; Calculate the actual stacking weight of the goods based on the actual stacking volume and standard weight of the goods; Step S36: Acquire shelf load limit data; perform shelf overload anomaly analysis based on the shelf load limit data and the actual stacking weight of the goods to obtain shelf overload anomaly data.

8. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 1 is characterized in that: Step S4 is specifically as follows: Step S41: performing fatigue detection on shelf connection nodes based on shelf overload abnormality data to obtain connection node fatigue data; Step S42: performing structural deformation detection based on the shelf overload abnormality data to obtain structural deformation data; Step S43: integrating the connection node fatigue data and the structural deformation data to obtain the shelf structure fatigue data; Step S44: predicting shelf life based on shelf structure fatigue data to obtain shelf life data; Step S45: Generate a shelf health heat map based on the shelf life data.

9. The method for intelligent warehouse digital visualization management based on the Internet of Things according to claim 8 is characterized in that: Step S41 is specifically as follows: Step S411: extracting the overload effect layer according to the shelf overload abnormality data to obtain the shelf overload data; Step S412: detecting the force conduction path based on the layer overload data to obtain the force conduction path data; Step S413: identifying stress concentration areas based on the force conduction path data; Step S414: locating high-risk laminate connection nodes based on stress concentration areas; Step S415: performing desoldering detection on high-risk layer connection nodes to obtain desoldering data; Step S416: determining the fatigue degree of the connection node according to the desoldering data, and obtaining the connection node fatigue data.

10. An intelligent warehouse digital visualization management system based on the Internet of Things, characterized by: The method for executing the intelligent warehouse digital visualization management method based on the Internet of Things according to claim 1, wherein the intelligent warehouse digital visualization management system based on the Internet of Things comprises: The material handling simulation module is used to obtain intelligent warehouse structure data; build a three-dimensional warehouse model based on the intelligent warehouse structure data; and perform material handling simulation based on the three-dimensional warehouse model to obtain material handling data; The cargo center of gravity offset detection module is used to identify high-frequency congestion areas based on material handling data; perform cargo placement analysis based on the high-frequency congestion areas to obtain cargo placement data; identify cargo spatial form based on the cargo placement data to obtain cargo spatial form data; and perform cargo center of gravity offset detection based on the cargo spatial form data to obtain cargo center of gravity offset data. The shelf overload anomaly analysis module is used to collect unique cargo labels based on cargo center of gravity offset data to obtain cargo label data; upload cargo label data to the warehouse management system and display the cargo storage status to obtain a real-time warehouse status diagram; perform shelf overload anomaly analysis based on the real-time warehouse status diagram to obtain shelf overload anomaly data; The shelf life prediction module is used to perform shelf structure fatigue detection based on shelf overload abnormal data to obtain shelf structure fatigue data; predict shelf life based on shelf structure fatigue data to obtain shelf life data; and generate a shelf health heat map based on the shelf life data.

Images