SVM Based Video Processing For Recognition And Tracking Of Pedestrians

The main research objective of this thesis is to detect and track pedestrian targets in video sequences.The detection and tracking of pedestrian targets has been a major research focus in recent years and has attracted widespread attention from many researchers.After researching and analyzing the traditional target recognition tracking algorithm,this thesis improves the related algorithm on this basis.The main research results of this thesis are as follows:(1)Improving the background model.After summarizing the advantages and disadvantages of the traditional algorithms,in order to improve the real-time performance of the detection algorithm,a background model with variable weights based on the average background model is proposed.The basic idea of the model is as follows.The background area of the image sequence is divided into four categories,and different weights are assigned to different background areas.From the experimental results,it can be seen that the model accurately reduces the foreground and background,and also reduces many redundant calculations,which improves the real-time performance of the model.(2)Gradient direction histogram(HOG)feature extraction based on information entropy.When extracting HOG features in local area,making full use of the distribution of the gradient direction,calculate the corresponding entropy value,and that value as the weight multiplied into the corresponding HOG feature vectors.Experiments show that the HOG feature vector combined with entropy can enhance the discrimination ability of SVM classifier.(3)The Support Vector Machine(SVM)algorithm detects the area of head portion of pedestrians.In practical situations,pedestrians will be affected by situations such as obstruction of body,different poses,or shooting angles,which makes it difficult for traditional pedestrian detection algorithms to extract pedestrians.In order to avoid the influence of the above factors on pedestrian detection,this thesis apply the SVM classification algorithm to detect the pedestrian head area.The detection of such a partial area of pedestrian's body not only avoid the influence of the above factors,but also reduce the volume of calculation during the detection and reduce the running time.