Design And Implementation Of Abnormal Behavior Detection System In Logistics Industrial Park

With the development of the Internet and e-commerce,China’s logistics industry is currently in a stage of vigorous development,but at the same time of rapid development,there are many safety issues that are easily overlooked by people,such as illegal use of mobile phones and smoking in industrial parks.Security issues arising from abnormal behavior.The combination of real-time monitoring and abnormal behavior detection algorithm is one of the ways to solve this problem,that is,to detect and record the abnormal behavior in the logistics park without human intervention.Traditional abnormal behavior detection algorithms based on mathematical methods often require a large time-space complexity,while abnormal behavior detection methods based on deep learning can easily learn the characteristics of abnormal behavior and identify abnormal behavior relatively quickly.However,most of the existing datasets only mark one abnormal behavior separately.If multiple abnormal behaviors need to be identified at the same time,the dataset needs to be re-labeled,which leads to an increase in workload and model identification delay.Therefore,this thesis proposes an integrated detection algorithm for abnormal behavior based on YOLO.This algorithm replaces the Backbone of YOLO with a lightweight neural network to improve the speed.Behavioral open source datasets are trained to improve its feature extraction capabilities,and finally the Backbone is shared,and the extracted features are sent to the Neck and Head parts of their respective YOLO for separate training and prediction.In order to verify the performance of the model,this thesis compares it with the currently popular abnormal behavior detection algorithms on four open source datasets.At the same time,this thesis designs and implements a set of abnormal behavior detection system with front-end and back-end separation.The system transmits the real-time video stream to the detection module.If abnormal behavior occurs,it will be marked according to the applied abnormal behavior detection algorithm and the key frame will be stored,which is convenient for later division of responsibilities and provides visual analysis of the stored abnormal behavior data function.The front end of the system is an HTML-based web application that provides users with a simple and easy-to-use interface.The back end uses the Flask framework to build web services to ensure the consistency of front-end and back-end data.Based on this,the development of user management,realtime detection,abnormal behavior data management and data visualization function modules has been completed.