Research On Intelligent Warehouse Management System Based On Deep Learning

With the widespread use of computer vision technology in industry,warehouse management has become an important area of application.One of the core technologies of intelligent warehousing is computer vision algorithms based on artificial intelligence,the application of which can effectively guarantee the quality of warehouse goods and warehouse safety.Through the collection of visual information generated during the warehousing operation,intelligent processing and analysis can effectively improve the efficiency and correctness of manual operations.Fibre and textile emergency relief materials are an important part of emergency relief materials,and the current warehouse management of its reserves generally uses manual inspection and video monitoring to complete the operation and maintenance work,which is inefficient and brings high manpower costs.In recent years,the rapid development of target recognition algorithms based on deep learning has provided technical guarantees for the realisation of product defect detection and fire detection in textile smart warehousing.Therefore,it is of great significance to reduce costs and increase efficiency in intelligent warehouse management by studying deep learningbased target detection algorithms and building textile intelligent warehouse management systems to achieve automatic detection of quality problems and safety issues.This paper selects the problems of fabric outbound defect detection and warehouse fire detection to carry out research on intelligent warehouse management systems based on deep learning,and solves the automation needs of quality management and safety management through computer vision technology.The main work of this paper is as follows:(1)A YOLOF-based fabric defect detection method is proposed for detecting fabric defect targets in complex backgrounds.The method replaces the multi-scale feature fusion architecture in YOLO by single-level feature detection,which significantly improves the detection efficiency and can be better adapted to practical operating situations in industrial environments.By comparing the YOLOF network with other methods,the experimental results show that the selected method has a higher average accuracy mean in fabric defect classification and can effectively improve the correct defect detection rate.(2)A Mobile Netv2-based warehouse fire detection method is proposed for achieving more accurate detection of flame targets in warehouses.The method improves the YOLOv3 network through the Mobile Netv2 module,which ensures lower information loss while reducing the amount of network computation.The improved YOLOv3 is compared with other models,and the experimental results show that the YOLOv3 network equipped with the Mobile Netv2 module can effectively improve the average accuracy and detection rate,effectively improving the ability to detect fires and reducing the risk of fires in warehouses,as well as reducing the rate of false fire alarms.(3)Research was conducted on an intelligent warehouse management system based on deep learning,including access management,intelligent operations and security control management.The YOLOF-based fabric defect detection and Mobile Netv2-based warehouse fire detection algorithms were deployed in the intelligent warehouse management system for practical application to verify the feasibility of the algorithms in this paper in the actual warehouse operation process,to achieve more efficient warehouse operations and to effectively solve the safety problems.