Design And Implementation Of Smart Court Electronic Foot Buckle Management System Based On Behavior Recognition

In recent years,there have been incidents of court detainees escaping on their way out of escort,which has aroused concern in the society and online media.Traditional shackles require constant supervision by the judicial police.This management method is likely to cause a waste of manpower and cannot realize the active prevention of dangerous behaviors.In response to the above drawbacks,it is urgent to establish a smart court electronic foot buckle system that combines the Internet of Things and artificial intelligence technology to actively prevent the escape of detainees and other abnormal behaviors,and realize the transition from passive control to proactive prevention and dynamic tracking.According to the needs of the smart court electronic foot buckle system,the system function modules include user management module,equipment management module,task management module,alarm management module,statistical analysis module,trajectory management module,map management module and behavior recognition module.In the user management module,use JWT to realize request login authentication;in the device management module,use the Netty gateway to realize the communication between the device and the server;use Quartz in the task management module to realize task pre-allocation push;use ECharts in the statistical analysis module to realize the display of charts;The management module uses an adaptive threshold to optimize the DP algorithm to achieve trajectory compression,and uses Redis and SSDB to query and store trajectory data;in the map management module,use Baidu Map API to achieve geofence setting and trajectory playback.In this article,the behavior recognition models constructed by the two methods are compared and analyzed.First,preprocess the data collected by the acceleration sensor,including data denoising and feature extraction.Then the principal component analysis method and linear discriminant analysis method commonly used in feature dimensionality reduction are analyzed,and a hybrid feature dimensionality reduction method is proposed to reduce the dimensionality of feature data.Then the performance of the three decision tree algorithms is analyzed through experiments,and the CART decision tree algorithm is used to construct a behavior recognition model after comparative analysis.The model can identify six behaviors:walking,running,sitting,standing,going upstairs,and going downstairs.Aiming at the insufficient classification ability of CART decision tree algorithm,the random forest algorithm is used to improve the accuracy of the improved model up to 87.92%.Finally,the improved algorithm is used to construct a behavior recognition model to assist police officers in identifying abnormal behaviors of prisoners.The system has finally been successfully trial-run,and has been affirmed by users of relevant units,and the system has certain reference significance for the development of smart courts in the Internet of Things and artificial intelligence environments.