Warehouse supervision-oriented AIoT edge computing intelligent analysis system and method

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a storage supervision-oriented AIoT edge computing intelligent analysis system and method, and the system comprises an edge computing node, a multi-modal data collection module, an intelligent analysis engine, a dynamic resource optimization module and a secure communication module. The system has the beneficial effects that real-time data acquisition and localized intelligent analysis of storage environments (temperature and humidity, illumination, gas concentration and the like) and cargo states (stacking integrity, displacement, damage and the like) are realized by integrating edge computing nodes and a multi-mode sensor; and the deep learning algorithm is used to carry out autonomous identification and early warning on abnormal events (such as cargo dumping, environment exceeding and illegal invasion).

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of artificial intelligence, and in particular to an AIoT edge computing intelligent analysis system and method for warehouse supervision. BACKGROUND

[0002] Traditional warehouse supervision systems mainly rely on manual inspection or centralized Internet of Things platforms, and have the following technical defects:

[0003] 1. Poor real-time performance: data needs to be uploaded to the cloud for processing, and network delays result in slow response, which cannot meet the immediate warning needs of high-risk scenarios (such as flammable product warehouses).

[0004] 2. High resource consumption: high-frequency data such as video streams are uploaded to the cloud throughout, occupying a large amount of bandwidth and computing resources.

[0005] 3. Low analysis accuracy: traditional algorithms based on rules have weak recognition ability for complex scenes (such as cargo stacking deformation and small leaks), and have high false positive and false negative rates.

[0006] 4. Poor scalability: the centralized architecture is difficult to support flexible deployment and collaborative analysis of a large number of sensor nodes. SUMMARY

[0007] The purpose of the present application is to provide an AIoT edge computing intelligent analysis system and method for warehouse supervision to solve the problems raised in the background.

[0008] To achieve the above purpose, the present application provides the following technical scheme: an AIoT edge computing intelligent analysis system for warehouse supervision, comprising:

[0009] An edge computing node is deployed in an embedded hardware in the warehouse, integrates a lightweight AI inference engine, and is used to support localized data processing and decision-making;

[0010] A multi-modal data acquisition module includes environmental sensors, visual sensors, mechanical sensors, RFID tag readers, and is used to realize synchronous acquisition of multi-dimensional data;

[0011] An intelligent analysis engine includes an anomaly detection model based on a lightweight target detection algorithm of an improved YOLOv8, which is used to identify cargo displacement, stacking abnormalities and illegal intrusion behaviors, and a time series prediction model using an LSTM network, which is used to analyze environmental data trends and predict over-standard risks;

[0012] A dynamic resource optimization module is used to dynamically allocate computing resources according to task priority and device remaining computing power;

[0013] A secure communication module supports encrypted data transmission and edge-cloud collaborative mechanisms, and backs up key data to a cloud blockchain for storage.

[0014] Preferably, the multi-modal data acquisition module, when implemented, deploys the device in key areas of the warehouse, and collects temperature and humidity, oxygen concentration data in real time through environmental sensors; captures images of goods stacking through visual sensors, and generates point cloud data in combination with 3D laser radar to detect the stacking inclination angle; monitors the shelf vibration frequency through mechanical sensors to judge the goods displacement risk; the data fusion module integrates multi-source data using Kalman filtering algorithm to generate a unified state matrix input into the intelligent analysis engine.

[0015] Preferably, when the intelligent analysis engine autonomously responds to abnormal events, when the detected goods stacking inclination angle exceeds the threshold value, a local alarm is triggered and alarm information is pushed to the administrator terminal; when the oxygen concentration is abnormally decreased, a gas leakage prediction model is started, and if the leakage risk is confirmed, the ventilation system is linked and the emergency team is notified; when the visual sensor identifies illegal intrusion behavior, face comparison and track tracking are started, and encrypted video stream is synchronized to the cloud for storage.

[0016] Preferably, the dynamic resource optimization module, in low power mode, reduces the camera resolution and switches to a lightweight model.

[0017] Preferably, the dynamic resource optimization module, when high concurrency tasks occur, enables model parallel computing, and distributes video stream analysis and environment prediction tasks to different computing units.

[0018] A method for an AIoT edge computing intelligent analysis system for warehouse supervision, comprising the following steps:

[0019] Through the multi-modal data acquisition module deployed in the key areas of the warehouse, the environmental sensors are used to collect temperature and humidity, gas, and smoke data in real time, the visual sensors capture images of goods stacking and generate point cloud data, the mechanical sensors monitor the shelf vibration frequency, and the RFID tag reader obtains the related information of the goods, realizing the synchronous collection of multi-dimensional data;

[0020] The data fusion module is used to integrate multi-source data using Kalman filtering algorithm to generate a unified state matrix;

[0021] The unified state matrix is input into the intelligent analysis engine, through the abnormality detection model based on the lightweight target detection algorithm of the improved YOLOv8, the goods displacement, stacking abnormality and illegal intrusion behavior are identified, and at the same time, the time series prediction model using LSTM network is used to analyze the environmental data trend and predict the risk of exceeding the standard.

[0022] Preferably, in the step of autonomous response to abnormal events:

[0023] When the intelligent analysis engine detects that the goods stacking inclination angle exceeds the preset threshold value, a local alarm device is triggered, and alarm information is pushed to the administrator terminal through the security communication module;

[0024] When the abnormal decrease in oxygen concentration is detected, a gas leakage prediction model is started, if the model confirms the presence of leakage risk, the ventilation system is opened through the safety communication module linkage, and the emergency team is notified;

[0025] When the visual sensor identifies illegal intrusion behavior, start face comparison and track tracking function, and at the same time, through the safety communication module, the encrypted video stream is synchronized to the cloud storage.

[0026] Preferably, it also includes a dynamic resource optimization step: according to the task priority and the remaining computing power of the device, the computing resources are dynamically allocated; in the low power mode, the camera resolution is reduced, and the lightweight model is switched to for data processing and analysis.

[0027] Preferably, in the dynamic resource optimization step, when facing high concurrent tasks, a model parallel computing mechanism is enabled, and video stream analysis and environment prediction tasks are allocated to different computing units for processing.

[0028] Preferably, the safety communication module supports encrypted data transmission during the entire method execution process, ensures the security of data during transmission, and realizes an edge-cloud collaborative mechanism to backup critical data to the cloud blockchain storage to ensure data integrity and non-tamperability.

[0029] Compared with the prior art, the beneficial effects of the present application are:

[0030] The AIoT edge computing intelligent analysis system and method for warehouse supervision provided by the present application realizes real-time data acquisition and localized intelligent analysis of warehouse environment (temperature and humidity, illumination, gas concentration, etc.) and goods state (stacking integrity, displacement, damage, etc.) by integrating edge computing nodes and multi-modal sensors; abnormal events (such as goods toppling, environmental over-limit, illegal intrusion) are autonomously identified and warned by using deep learning algorithms; the edge computing efficiency is optimized by combining dynamic resource allocation technology, reducing cloud dependence and network delay. The present application solves the problems of data processing lag, high resource consumption of centralized system, high false alarm and missed alarm rate, etc. in traditional warehouse supervision, has the characteristics of strong real-time, high analysis accuracy, and high resource utilization, and can be widely applied in intelligent warehouse, cold chain logistics, dangerous goods storage and other fields. BRIEF DESCRIPTION OF DRAWINGS

[0031] Fig. 1 The system architecture diagram of the present application;

[0032] Fig. 2 The multi-modal data fusion flowchart of the present application;

[0033] Fig. 3 The dynamic resource allocation logic diagram of the present application;

[0034] Fig. 4 Hardware deployment schematic for the present application. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution of the present application, and the advantages more clear, the embodiments of the present application are further described in detail below in combination with the drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present application, not all embodiments, and are only used to explain the embodiments of the present application, and do not limit the embodiments of the present application. All other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0036] Embodiment one, please refer to Figs. 1 to 4 The present application provides a technical solution: an AIoT edge computing intelligent analysis system for warehouse supervision, comprising:

[0037] 1. Edge computing node: embedded hardware deployed in the warehouse, integrating lightweight AI inference engine, supporting localized data processing and decision-making.

[0038] 2. Multi-modal data acquisition module: including environmental sensors (temperature and humidity, gas, smoke), visual sensors (infrared camera, 3D laser radar), mechanical sensors (pressure, vibration) and RFID tag reader, realizing multi-dimensional data synchronous acquisition.

[0039] 3. Intelligent analysis engine:

[0040] Anomaly detection model: lightweight target detection algorithm based on improved YOLOv8, identifying cargo displacement, stacking abnormalities and illegal intrusion behavior

[0041] Time series prediction model: using LSTM network to analyze environmental data trends and predict over-limit risks (such as temperature and humidity deviation from threshold).

[0042] 4. Dynamic resource optimization module: dynamically allocating computing resources (such as video stream resolution degradation, model quantization acceleration) according to task priority and device remaining computing power.

[0043] 5. Secure communication module: supporting encrypted data transmission and edge-cloud collaboration mechanism, backing up critical data to cloud blockchain for storage.

[0044] Embodiment two, on the basis of embodiment one, multi-modal data acquisition and fusion are proposed:

[0045] 1. The device is deployed in the key area of the warehouse, and real-time acquisition of temperature and humidity, oxygen concentration data is realized through the environmental sensor.

[0046] 2. Visual sensor captures goods stack image, generates point cloud data combined with 3D laser radar, detects stack inclination angle.

[0047] 3. Mechanical sensor monitors shelf vibration frequency, judges goods displacement risk.

[0048] 4. Data fusion module integrates multi-source data using Kalman filtering algorithm, generates unified state matrix input intelligent analysis engine.

[0049] Example three, on the basis of example two, proposes an abnormal event autonomous response:

[0050] 1. Intelligent analysis engine detects that the goods stack inclination angle exceeds the threshold, triggers local alarm and pushes alarm information to administrator terminal.

[0051] 2. Detecting abnormal decrease in oxygen concentration, starting gas leakage prediction model, if confirming leakage risk, linking ventilation system and notifying emergency team.

[0052] 3. Visual sensor identifies illegal intrusion behavior, starts face comparison and track tracking, synchronously encrypts video stream to cloud storage.

[0053] Example four, on the basis of example three, proposes dynamic resource optimization:

[0054] 1. In low power mode, dynamic resource optimization module reduces camera resolution and switches to lightweight model (MobileNet-YOLO).

[0055] 2. When high concurrency task, enable model parallel computing, assign video stream analysis and environment prediction task to different computing units.

[0056] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An AIoT edge computing intelligent analysis system for warehouse supervision, characterized by: include: Edge computing nodes are embedded hardware deployed in warehouses and integrate lightweight AI inference engines to support localized data processing and decision-making. Multimodal data acquisition module, including environmental sensors, visual sensors, mechanical sensors and RFID tag readers, is used to achieve synchronous multi-dimensional data acquisition; An intelligent analysis engine, including an anomaly detection model based on an improved YOLOv8 lightweight object detection algorithm, used to identify cargo displacement, stacking anomalies, and illegal intrusions; and a time series prediction model using an LSTM network to analyze environmental data trends and predict the risk of exceeding standards; Dynamic resource optimization module, used to dynamically allocate computing resources based on task priority and remaining computing power of the device; The secure communication module supports encrypted data transmission and edge-cloud collaboration, and backs up key data to the cloud blockchain for storage.

2. The AIoT edge computing intelligent analysis system for warehouse supervision according to claim 1 is characterized by: When implementing the multimodal data acquisition module, devices are deployed in key areas of the warehouse. Environmental sensors collect real-time temperature, humidity, and oxygen concentration data. Visual sensors capture images of stacked goods and combine them with 3D lidar to generate point cloud data to detect stack tilt angles. Mechanical sensors monitor shelf vibration frequency to determine the risk of cargo displacement. The data fusion module uses the Kalman filter algorithm to integrate multi-source data and generate a unified state matrix to input into the intelligent analysis engine.

3. The AIoT edge computing intelligent analysis system for warehouse supervision according to claim 2 is characterized by: When the intelligent analysis engine autonomously responds to abnormal events, if it detects that the tilt angle of the cargo stack exceeds the threshold, it triggers a local alarm and pushes the alarm information to the administrator terminal; when it detects an abnormal drop in oxygen concentration, it activates the gas leakage prediction model. If the leakage risk is confirmed, the ventilation system is linked and the emergency team is notified; when the visual sensor identifies illegal intrusion, it activates facial comparison and trajectory tracking, and synchronizes the encrypted video stream to the cloud for evidence storage.

4. The AIoT edge computing intelligent analysis system for warehouse supervision according to claim 3 is characterized by: The dynamic resource optimization module reduces the camera resolution and switches to a lightweight model in low power mode.

5. The AIoT edge computing intelligent analysis system for warehouse supervision according to claim 4 is characterized by: The dynamic resource optimization module enables model parallel computing when performing high-concurrency tasks, and distributes video stream analysis and environmental prediction tasks to different computing units.

6. A method for an AIoT edge computing intelligent analysis system for warehouse supervision according to claim 5, characterized in that: The following steps are involved: Multimodal data acquisition modules deployed in key areas of the warehouse use environmental sensors to collect temperature, humidity, gas, and smoke data in real time. Visual sensors capture images of stacked goods and generate point cloud data. Mechanical sensors monitor shelf vibration frequency. RFID tag readers and writers acquire cargo-related information, enabling simultaneous multi-dimensional data collection. Using the data fusion module, the Kalman filter algorithm is used to integrate multi-source data and generate a unified state matrix; The unified state matrix is ​​input into the intelligent analysis engine, and the anomaly detection model based on the improved YOLOv8 lightweight target detection algorithm is used to identify cargo displacement, stacking anomalies and illegal intrusion behaviors. At the same time, the time series prediction model using the LSTM network is used to analyze environmental data trends and predict the risk of exceeding the standard.

7. A method according to claim 6, characterized in that: In the autonomous response step of abnormal events: When the intelligent analysis engine detects that the tilt angle of the cargo stack exceeds the preset threshold, it triggers the local alarm device and pushes the alarm information to the administrator terminal through the secure communication module; When an abnormal drop in oxygen concentration is detected, the gas leakage prediction model is activated. If the model confirms the risk of leakage, the ventilation system is activated through the safety communication module and the emergency team is notified. When the visual sensor identifies illegal intrusion, it activates the face matching and trajectory tracking functions, and synchronizes the encrypted video stream to the cloud for evidence storage through the secure communication module.

8. A method according to claim 7, characterized in that: It also includes dynamic resource optimization steps: dynamically allocating computing resources based on task priority and the remaining computing power of the device; in low-power mode, reducing the camera resolution and switching to a lightweight model for data processing and analysis.

9. A method according to claim 8, characterized in that: In the dynamic resource optimization step, when faced with high-concurrency tasks, the model parallel computing mechanism is enabled to assign video stream analysis and environment prediction tasks to different computing units for processing.

10. A method according to claim 9, characterized in that: The secure communication module supports encrypted data transmission during the entire method execution process to ensure the security of data during transmission. At the same time, it realizes the edge-cloud collaboration mechanism and backs up key data to the cloud blockchain for storage to ensure data integrity and non-tamperability.

Images