Research On People Detection And Analysis Base On Video Image

Due to the continuous development of the social economy and more and morepeople’s social activities, the requirement of security has become more and moreurgent, video surveillance system began to be widely used in many ways. It is adevelopment trend to use the computer vision to improve the automation of videosurveillance system and reduce human intervention in the future. Using the videosurveillance system as a platform, it is of extremely significance to get the number ofthe pedestrians who are detected and tracked in the video sequences. Recently, thestudy of people counting by video surveillance technology has become a hot anddifficult field of computer vision.This article related image processing algorithms and existing moving targetdetection and tracking algorithms for the analysis by video surveillance, designing anumber of video surveillance statistical algorithm flow. The people countingalgorithm was mainly divided into four parts which involved the imagepre-processing, moving target detection, target tracking and trajectory analysis. Inview of the current problems in several parts, this paper presents a number ofprograms and improved methods to achieve the crowd video image detection andcounting.Moving targets detection was the basis of the people counting system in thethesis, first, for an ordinary video images, comparing broad application of severalalgorithms’ advantages and disadvantages, since the application environments of thevideo surveillance comprised stationary background, we exploited backgroundsubtraction approach to extract the moving targets. This paper focused on Gaussianmixture background model, and for its disadvantage such as computing capacity, falsedetection, This paper presents an improved method to combine a three-framedifference method with the Gaussian mixture background model, moving foregroundedge information obtained relatively complete preservation, which helps to deal withfollow-up work.In the aspect of moving targets tracking, analysis of several common trackingalgorithm, In view of each algorithm’s disadvantage such as wrong tracking、 tracking lost or large amount of calculation, etc, this paper presents an object trackingalgorithm pedestrian movement information with a combination of Kalman Filtering,the algorithm uses information do pedestrian movement tracking, it can be a goodsolution to a variety of situations, such as pedestrian movement appeared cross,overlapping, separately,etc.After the realization of pedestrians tracking, this thesis analyzed the trajectory ofthe target and presented an effective people counting method.Real-time performance and practicality was the goal of this thesis, a number ofkey technologies in people counting in video surveillance were studied and a robustmethod with low complexity was proposed, which achieved good results in theexperiment.