Research On Application Of Human Abnormal Behavior Identification In Surveillance System

In the last few decades,rapid development of computer technology is widely recognized.Follow this is the rising pressure applied to intelligent monitoring systems for detecting abnormal events.It is difficult to meet the requirement of monitoring multiple devices at such a large quantity.What's more,it would cause the fatigue for security personnel.Comparing with normal behavior,the probability of abnormal behavior is low,which makes the task of monitoring more challenging and complex.The video system,which equipped with the human body abnormal behavior recognition and monitoring function,can monitor and identify whether people in the abnormal conditions in real time.When an abnormal event occurs,this system can be processed and handle in time.Meanwhile,it can achieve resource conservation and optimal results.Therefore,comparing with traditional surveillance system,the intelligent video surveillance system,with its advantages in cost and the economic value and its huge market demand,attracted the attention of scholars and institutions worldwide.There are many differences between the intelligent video surveillance system and the traditional video surveillance system.The primary difference is that the former has the ability to handle problems autonomously and can handle the emergency in a timely manner.In this way,the computer can assist humans and even replace humans to monitor in public places.Eventually,surveillance workload can be reduced significantly.This new processing approach is of great research value as a video monitoring,surveillance scene modeling and image retrieval process.Based on some existing algorithms and theories in this field,this paper expounded the key points of the detection of moving targets,the tracking of human targets and the judgement of human anomalous behaviors.The main work and details are as follows:(1)Introducing the working principles and experimental methods of some common moving object detection algorithms.The commonly used methods of foreground detection technology are presented: the optical flow method,neighboring frame subtraction,background subtraction and edge detection method.Then this paper briefly described their algorithm principle,advantages and disadvantages.After that,the paper is focused on the study of background subtraction.The background subtraction is then performed using Vibe algorithm modelling to show the core ideas,workflow and conducted experiments.(2)Introducing the technique of human target recognition and tracking.Firstly,this paper introduced the method of identifying whether the moving target is a human target based on human characteristic;Secondly,it presented the ideas and operation steps of MeanShift algorithm;Finally,using MeanShift algorithm and Kalman filter,the effect on the moving target tracking is tested.(3)Defining the classification of abnormal behavior.Afterwards,this paper showed the adopted different conditions.Finally,by comparing the motion characteristics and conditions of the processed images,it can be judged and verified whether it is an abnormal behavior or not.It is verified by simulation experiments.(4)Introducing the experiment platform,environment configuration and overall structure flow which used in this experiment.In this paper,the dynamic target recognition algorithm,human target tracking algorithm and related image processing are all experimented and tested on the Visual Studio 2013 platform and MATLAB 2015.Finally,through the processing of self-captured video clips and video data in the video library,it is verified that the treatment method used for human abnormal behavior recognition and related algorithms used in this paper have good experimental results and feasibility.