An AI-based warehouse intelligent supervision management method
By deploying AI technology in the warehouse to collect and analyze real-time environmental data, identify and warn of security risks, the problem of lack of real-time early warning in traditional warehouse management methods is solved, and more efficient security management is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA COMMUNICATIONS SUPPLY CHAIN CO LTD SICHUAN BRANCH
- Filing Date
- 2026-01-07
- Publication Date
- 2026-06-02
Smart Images
Figure CN122134236A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of warehouse supervision and management technology, and in particular to an AI-based intelligent warehouse supervision and management method. Background Technology
[0002] Warehouse management, especially for medium and large-sized warehousing and logistics centers, is a core link in the modern supply chain. The safety and efficiency of its operation directly affect the company's operating costs and service quality.
[0003] Traditional warehouse safety and management supervision methods involve deploying surveillance cameras in key areas of the warehouse to record videos. However, this method can only be used for post-incident tracing and evidence collection. That is, after an accident occurs, managers can review the video to find out the cause and responsibility. It lacks the ability to proactively identify and provide real-time early warnings, and cannot effectively intervene when risks occur or violations occur, resulting in poor warehouse safety management. Summary of the Invention
[0004] The purpose of this invention is to provide an AI-based intelligent supervision and management method for warehouses, which can intelligently supervise and manage warehouses, effectively intervene when risks occur or violations occur, and improve the safety management of warehouses.
[0005] To achieve the above objectives, the present invention provides an AI-based intelligent warehouse supervision and management method, comprising: Collect real-time environmental data from the warehouse; The collected real-time environmental data is transmitted to edge computing nodes for preprocessing and real-time analysis. Multiple artificial intelligence analysis models are run on edge computing nodes to process real-time environmental data in parallel and identify security risk behaviors in the warehouse; When a security risk behavior is detected and reaches the warning threshold, a tiered warning message is generated. The corresponding real-time intervention mechanism is triggered based on the warning level.
[0006] The process of triggering the corresponding real-time intervention mechanism based on the warning level also includes: All early warning events and intervention records are uploaded to a cloud management platform for storage and analysis.
[0007] The specific steps for collecting real-time environmental data from the warehouse include: Video streams and image data of the warehouse monitoring area are acquired through visual acquisition devices deployed in the warehouse; The precise location information of personnel, forklifts, or AGVs is obtained in real time through positioning devices deployed in the warehouse. Data on warehouse temperature, humidity, smoke concentration, and equipment vibration are collected using environmental monitoring equipment deployed in the warehouse.
[0008] The specific steps for transmitting the collected real-time environmental data to edge computing nodes for preprocessing and real-time analysis include: The received video stream data is decoded and frames are extracted at the edge computing nodes; Denoising, enhancement, and standardization preprocessing are performed on the image data; Data collected by positioning and environmental monitoring equipment is cleaned, filtered, and spatiotemporally aligned to provide a structured data input for artificial intelligence analysis models.
[0009] The specific steps involved in running multiple AI analysis models on edge computing nodes to process real-time environmental data in parallel and identify security risk behaviors in the warehouse include: The behavior recognition model, object detection model, trajectory prediction model, and PPE recognition model run in parallel on the edge computing node. The behavior recognition model identifies abnormal behaviors such as climbing, running, and falling based on real-time environmental data. The object detection model detects blocked fire exits and improperly stacked goods based on real-time environmental data. The trajectory prediction model predicts the collision risk between mobile devices based on real-time environmental data. The PPE recognition model detects whether safety protective equipment is being worn based on real-time environmental data.
[0010] The specific steps for generating tiered early warning information when a security risk behavior is identified and reaches the warning threshold include: Preset multi-level early warning thresholds corresponding to different safety risk behaviors; The risk behavior characteristics identified by multiple artificial intelligence analysis models are compared with the warning threshold. Based on the comparison results, tiered early warning information is generated.
[0011] Among them, in the step of generating graded early warning information based on the comparison results, The early warning information includes the risk type, risk location, and risk level.
[0012] This invention provides an AI-based intelligent warehouse supervision and management method. Through an artificial intelligence analysis model, it continuously analyzes the warehouse environment and can issue early warnings at the onset or even before safety risks occur, while triggering corresponding real-time intervention mechanisms to stop incidents before they happen. This greatly improves the initiative and timeliness of safety management. This invention enables intelligent supervision and management of warehouses, allowing for effective intervention when risks occur or violations occur, thereby improving the effectiveness of warehouse safety management. Attached Figure Description
[0013] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.
[0014] Figure 1 This is a flowchart of an AI-based intelligent warehouse supervision and management method according to the present invention.
[0015] Figure 2 This is a flowchart of the real-time environmental data collection in the warehouse of this invention.
[0016] Figure 3 This is a flowchart of the present invention, which transmits the collected real-time environmental data to an edge computing node for preprocessing and real-time analysis.
[0017] Figure 4 This invention provides a flowchart for generating tiered early warning information when a security risk behavior is detected and reaches an early warning threshold.
[0018] Figure 5 This is a flowchart of the present invention that uploads all early warning events and intervention records to a cloud management platform for storage and analysis. Detailed Implementation
[0019] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, but should not be construed as limiting the present invention.
[0020] Please see Figures 1-5 This invention provides an AI-based intelligent warehouse supervision and management method, comprising: S1 collects real-time environmental data from the warehouse; The specific steps include: S11 acquires video streams and image data of the warehouse monitoring area through visual acquisition devices deployed in the warehouse; In this embodiment of the invention, high-definition network cameras are deployed in key areas such as warehouse entrances, aisles, high-bay racking areas, loading and unloading platforms, and forklift charging areas to obtain global monitoring video streams; in accident-prone areas, such as corners of racking aisles and entrances and exits, AI edge computing cameras with built-in lightweight AI algorithms are deployed for initial real-time image acquisition.
[0021] S12 obtains the precise location information of personnel, forklifts, or AGVs in real time through positioning devices deployed in the warehouse; In this embodiment of the invention, all forklifts, AGVs, and warehouse operators are equipped with UWB high-precision positioning tags. By deploying UWB positioning base stations on the top of the warehouse, centimeter-level real-time tracking of the dynamic positions of personnel and equipment is achieved.
[0022] S13 collects data on warehouse temperature, humidity, smoke concentration, and equipment vibration through environmental monitoring equipment deployed in the warehouse; In this embodiment of the invention, temperature and humidity sensors, smoke sensors, and vibration sensors are deployed on the top of the warehouse, on the shelves, and next to the fire exit to collect environmental safety data.
[0023] S2 transmits the collected real-time environmental data to edge computing nodes for preprocessing and real-time analysis; The specific steps include: S21 decodes and extracts frames from the received video stream data at the edge computing node; S22 performs noise reduction, enhancement, and standardization preprocessing on the image data; S23 performs data cleaning, filtering, and spatiotemporal alignment operations on the data collected by positioning devices and environmental monitoring devices, providing a regular data input for the analysis of artificial intelligence analysis models; In this embodiment of the invention, an edge computing server is deployed in a warehouse server room or on-site cabinet. All collected real-time environmental data is aggregated to the edge server through the industrial switch network inside the warehouse. The edge server decodes the received video stream data and extracts frames at a preset frequency to standardize the image resolution to 1080P. The extracted image frames are then subjected to Gaussian denoising and histogram equalization to improve image quality. Data from UWB base stations and environmental sensors such as temperature and humidity sensors, smoke sensors, and vibration sensors are cleaned and Kalman filtered, and data from different sources are aligned and fused in time and space.
[0024] S3 runs multiple artificial intelligence analysis models on edge computing nodes to process real-time environmental data in parallel and identify security risk behaviors in the warehouse; The specific steps include: running a behavior recognition model, an object detection model, a trajectory prediction model, and a PPE recognition model in parallel on edge computing nodes. The behavior recognition model identifies abnormal behaviors such as climbing, running, and falling based on real-time environmental data. The object detection model detects blocked fire exits and improperly stacked goods based on real-time environmental data. The trajectory prediction model predicts the collision risk between mobile devices based on real-time environmental data. The PPE recognition model detects the wearing of safety protective equipment based on real-time environmental data.
[0025] In this embodiment of the invention, multiple pre-trained artificial intelligence analysis models are run in parallel on an edge server using its GPU computing power. These models include a behavior recognition model, an object detection model, a trajectory prediction model, and a PPE recognition model. The behavior recognition model, developed based on the existing YOLOv7-Pose pose estimation model, is used to continuously analyze video streams and identify abnormal behaviors such as "people climbing shelves," "running quickly," and "sudden falls" in real time. The object detection model, developed based on the existing YOLOv8-seg instance segmentation model, is used to continuously monitor areas such as fire exits and emergency exits, identifying whether goods, pallets, or equipment have been placed and remained stationary for more than a preset time. The trajectory prediction model, developed based on the existing Social-GAN social generative adversarial network, receives fused UWB positioning data and calculates the relative distance, speed, and direction of movement between forklifts and personnel, and between forklifts, in real time, predicting the probability of collisions within the next 2-5 seconds. The PE (Personal Protective Equipment) identification model, developed based on the existing YOLOv8 model, is used to automatically call the camera in the preset "high shelf area" when the system determines that a person has entered the area through UWB positioning, and identify whether the person is wearing a safety helmet and reflective clothing.
[0026] When S4 detects a security risk behavior that reaches the warning threshold, it generates a tiered warning message. The specific steps include: S41 presets multi-level early warning thresholds corresponding to different safety risk behaviors; S42 compares the risk behavior characteristics identified by multiple artificial intelligence analysis models with the warning threshold; Based on the comparison results, S43 generates graded early warning information.
[0027] In this step, the warning information includes the risk type, risk location, and risk level.
[0028] In this embodiment of the invention, multiple warning thresholds are preset. For example, a predicted collision probability > 70% is a Level 3 warning; a fire lane blockage for > 5 minutes is a Level 2 warning; and entering a specific area without wearing a safety helmet is a Level 1 warning. The artificial intelligence analysis model compares the identification results (e.g., "Entrance to Lane 3, collision probability 85%)" with the preset thresholds. If the comparison is successful, it automatically generates structured hierarchical warning information. This level of warning information is a JSON data packet containing the following fields: risk type (e.g., "collision risk", "PPE violation"), risk location (e.g., "Area A - Lane 2"), risk level (e.g., "Level 3"), timestamp, and suggested measures (e.g., "immediately slow down and avoid").
[0029] S5 triggers the corresponding real-time intervention mechanism based on the warning level; In this embodiment of the invention, upon receiving an early warning message, a preset intervention script is triggered based on its risk level field. For example, a Level 1 warning triggers an IP voice broadcast loudspeaker near the risk area to play a pre-recorded reminder message, such as "You have entered the high-bay racking area, please wear your safety helmet." A Level 2 warning pushes an alarm notification to the PDA or mobile APP of the administrator responsible for the area, and simultaneously triggers the on-site audible and visual alarms to flash yellow. A Level 3 warning triggers high-decibel audible and visual alarms throughout the warehouse to flash red and sound. At the same time, the system sends a command to the controller of the forklift involved through the IoT gateway to forcibly limit its maximum speed until the risk is eliminated.
[0030] S6 uploads all early warning events and intervention records to a cloud management platform for storage and analysis; The specific steps include: S61 uploads data packets related to the warning event, including timestamps, location information, video evidence, warning level, and intervention measures, to the cloud management platform; S62 establishes a security incident database on a cloud management platform for long-term trend analysis, report generation, and iterative optimization of artificial intelligence analysis models.
[0031] In this embodiment of the invention, the edge server encrypts and uploads the complete data packet for each early warning event, including video clips before and after the trigger, sensor data, early warning information, executed intervention measures, and results, to the cloud management platform via HTTPS. The cloud platform stores the data in a time-series database and establishes a security event knowledge graph. The platform utilizes the accumulated data for macro-level analysis, such as using clustering algorithms to identify high-incidence times and locations of incidents; using association rule analysis to identify the preconditions for specific violations; generating weekly and monthly security management reports; and providing decision support dashboards for managers. This data is also used for incremental training and version iteration of the AI analysis model deployed at the edge, thereby achieving system self-optimization and continuous evolution.
[0032] This invention provides an AI-based intelligent warehouse supervision and management method that continuously analyzes the warehouse environment using behavior recognition models, object detection models, trajectory prediction models, and PPE recognition models. It can issue early warnings at the onset or even before safety risks occur, triggering corresponding real-time intervention mechanisms to prevent incidents before they happen, thus greatly improving the initiative and timeliness of safety management. This invention enables intelligent supervision and management of warehouses, allowing for effective intervention at the moment risks occur or violations occur, thereby improving the effectiveness of warehouse safety management.
[0033] The above-disclosed embodiments are merely one or more preferred embodiments of this application and should not be construed as limiting the scope of this application. Those skilled in the art can understand that all or part of the processes for implementing the above embodiments and equivalent changes made in accordance with the claims of this application still fall within the scope of this application.
Claims
1. An AI-based intelligent warehouse supervision and management method, characterized in that, include: Collect real-time environmental data from the warehouse; The collected real-time environmental data is transmitted to edge computing nodes for preprocessing and real-time analysis. Multiple artificial intelligence analysis models are run on edge computing nodes to process real-time environmental data in parallel and identify security risk behaviors in the warehouse; When a security risk behavior is detected and reaches the warning threshold, a tiered warning message is generated. The corresponding real-time intervention mechanism is triggered based on the warning level.
2. The AI-based intelligent warehouse supervision and management method as described in claim 1, characterized in that, After triggering the corresponding real-time intervention mechanism based on the warning level, the steps also include: All early warning events and intervention records are uploaded to a cloud management platform for storage and analysis.
3. The AI-based intelligent warehouse supervision and management method as described in claim 2, characterized in that, The specific steps for collecting real-time environmental data from the warehouse include: Video streams and image data of the warehouse monitoring area are acquired through visual acquisition devices deployed in the warehouse; The precise location information of personnel, forklifts, or AGVs is obtained in real time through positioning devices deployed in the warehouse. Data on warehouse temperature, humidity, smoke concentration, and equipment vibration are collected using environmental monitoring equipment deployed in the warehouse.
4. The AI-based intelligent warehouse supervision and management method as described in claim 3, characterized in that, The specific steps for transmitting the collected real-time environmental data to edge computing nodes for preprocessing and real-time analysis include: The received video stream data is decoded and frames are extracted at the edge computing nodes; Denoising, enhancement, and standardization preprocessing are performed on the image data; Data collected by positioning and environmental monitoring equipment is cleaned, filtered, and spatiotemporally aligned to provide a structured data input for artificial intelligence analysis models.
5. The AI-based intelligent warehouse supervision and management method as described in claim 4, characterized in that, Running multiple AI analytics models on edge computing nodes to process real-time environmental data in parallel and identify security risk behaviors in the warehouse involves the following steps: The behavior recognition model, object detection model, trajectory prediction model, and PPE recognition model run in parallel on the edge computing node. The behavior recognition model identifies abnormal behaviors such as climbing, running, and falling based on real-time environmental data. The object detection model detects blocked fire exits and improperly stacked goods based on real-time environmental data. The trajectory prediction model predicts the collision risk between mobile devices based on real-time environmental data. The PPE recognition model detects the wearing of safety protective equipment based on real-time environmental data.
6. The AI-based intelligent warehouse supervision and management method as described in claim 5, characterized in that, When a security risk behavior is identified and reaches the warning threshold, the specific steps for generating a tiered warning message include: Preset multi-level early warning thresholds corresponding to different safety risk behaviors; The risk behavior characteristics identified by multiple artificial intelligence analysis models are compared with the warning threshold. Based on the comparison results, tiered early warning information is generated.
7. The AI-based intelligent warehouse supervision and management method as described in claim 6, characterized in that, Based on the comparison results, in the step of generating tiered early warning information... The early warning information includes the risk type, risk location, and risk level.