Research On Moving Target Detection And Tracking Technology In Video Surveillance

With the rapid development of information construction video surveillance has become very popular.Video surveillance is an important information carrier for human beings because of recording a lot of information and a wide range of monitoring.Moving object detection and tracking technology in video surveillance becomes the focus of people's attention with the continuous development of artificial intelligence technology.The traditional methods of detecting and tracking targets rely on human beings are prone to miss detection,which requires researchers to use machine vision technology and artificial intelligence technology to conduct in-depth research on video monitoring and identify,track and analyze monitoring information accurately.The research mainly studies the moving target detection and tracking technology in surveillance video aiming to achieve the target detection and tracking task in surveillance video.(1)It introduces the technology of moving target detection and tracking in video surveillance,and makes a systematic analysis of the present monitoring system,and points out the necessity of applying intelligent monitoring in monitoring.(2)The research focus the methods of image preprocessing.The algorithm of self-adaptive median filter has better effect in image denoising compared with algorithm of Gauss filter,median filter and adaptive median filter in the image denoising algorithm.The method of moving object detection is researched in optical flow method,frame difference method and background difference method.An improved Gaussian mixture model is designed by adding three frame difference algorithm to deal with the fuzzy edge of moving object,which can detect the edge of moving object well.The seed filling method is used to remove the void phenomenon in the result of detection.The method of HOG+SVM is used to detect the pedestrian in the video surveillance and good results have been achieved.(3)The principle and structure of convolutional neural network are studied,and the SiamFC network model is analyzed.The tracking of moving target in SiamFC network is realized.The research proposes a target tracking algorithm based on feature fusion by comparing the feature extraction characteristics of neural network in background interference.The test is carried out on the data of OTB100 through the FPN network integration of layer 4 and layer 5 features,and the success rate of tracking AUC has some improvement in the case of background interference.(4)We design and develop a video surveillance moving object detection and tracking system based on the deep learning training model,which can effectively detect and track the moving objects in the video.