Research On Pedestrian Detection And Tracking In Indoor Surveillance Video

Nowadays,with the development of the social economy and the improvement of the public safety awareness,a large number of video surveillance systems are installed not only in shopping malls and train stations,but also in private homes.We wish that the surveillance systems would provide more information about the pedestrians' behaviors.Pedestrian detection and tracking are the base of that and widely used in intelligent traffic systems,HCI and safety monitoring.Classic pedestrian detection and tracking algorithms are described and their advantages and disadvantages are analyzed in the paper.Therefore,we put forward a new system of pedestrian detection and tracking in indoor surveillance video.The main works are as follows:Firstly,this paper describes the extraction processing of the HOG feature in details.According to the fact that the dimension of HOG feature is high and to calculate the classification result takes a long time,we extracts HOG feature based on PCA algorithm and integral image is introduced.A new method of pedestrian detection is proposed based on the HOG feature and SVM classifier.Secondly,aimed at the background of the indoor surveillance video,we use Gaussian Mixed Model to get the moving regions.The comparative experiment on Gaussian Mixed Model and Background Difference is analyzed.Combined with the moving regions detection,the pedestrian detection algorithm runs faster and more efficiently.Thirdly,Camshift algorithm is analyzed in details.According to its disadvantages,an improved algorithm is proposed based on Camshift algorithm and moving regions detection.At last,a system of pedestrian detection and tracking in indoor surveillance video is designed.