Research And Implementation Of Moving Object Detection Algorithm In Intelligent Monitoring System

In recent years,with the development of the economic level,people's safety awareness has been continuously improved,and security needs in various fields of society have continuously increased.At the same time,computer technology and Internet technology have developed rapidly.Market demand and technological development have made the intelligent video surveillance system a research hotspot.At present,intelligent video surveillance systems have been used in many areas such as home protection,banking,stations,airports,and defense to completely solve the inefficiency and unreliability of traditional video surveillance systems relying on human analysis of video content.The detection and tracking of moving targets is an important application area of intelligent monitoring systems.This paper constructs an embedded network intelligent video monitoring system based on ARM by researching and improving the moving target detection algorithm and shadow removal algorithm.The main work of this article is as follows:First,for the detection of moving targets this paper proposes a method that combines continuous three-frame difference method with background subtraction.Firstly,the video frame sequence is processed successively with three frames to determine the moment of the moving target.The video frame immediately before the invasion of the moving target is used as the initial template of the background model.After the background model is established,the moving target is extracted by the background subtraction,and then the background is updated.The non-moving target portion in the video frame replaces the corresponding region in the background model,and the other regions remain unchanged.Finally,we combine the set of moving targets obtained by the background subtraction method and the three-frame difference method to determine the final moving target and avoid the "void" phenomenon in the single frame difference method.Second,due to the current moving target detection method can perform motion detection but cannot distinguish between foreground target and moving shadow,it has great influence on the target tracking in the next stage.This paper proposes a method based on color features,normalized vector distance,and brightness ratio to remove moving shadow.First,a background image is created by mixing the Gaussian model,and the motion region is separated by the background subtraction method.Then,the serial processing method is used to detect the shadow pixels in the motion area.After the shadow is removed according to the color consistency feature in the RGB color space,the shadow pixels are further detected according to the normalized vector distance distribution histogram of the motion area.Finally,for the problem of mistaken detection in the shadow detection process,a pixel illumination model is established to calculate the ratio of the brightness of the shadow pixel to the background pixel,and the false detection of foreground pixels is eliminated based on the confidence interval.Third,an embedded network video surveillance system based on ARM is implemented.This article uses S3C2440 chip as the hardware core,Linux system as the operating system,adopts B/S architecture(browser/server)architecture,combined with the moving target detection algorithm,completes the design and implementation of video data collection,compression,motion detection and mail alarming.Experimental results show that the ARM-based embedded intelligent network monitoring system constructed in this paper can effectively capture and display video data,detect moving targets,and perform email alerts.