Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Study of Pedestrian and Unusual Activities Detection Methods Based on Video Sequence

Author CuiGuoQing
Tutor ZengGuiHua
School Shanghai Jiaotong University
Course Communication and Information System
Keywords Pedestrian Detection Abnormal behavior detection Gradient direction histogram AdaBoost algorithm MILBoost algorithm Lucas-Kanade optical flow method
CLC TP391.41
Type Master's thesis
Year 2010
Downloads 616
Quotes 2
Download Dissertation

Pedestrian Detection with abnormal behavior detection in daily intelligent transportation, intelligent monitoring , multimedia search field has a very wide range of applications . Based on the technology , we found that the occurrence of abnormal behavior the pedestrians and monitoring scenes video . However, due to the diversity of monitoring changes in the scene and the pedestrian posture , pedestrian detection and abnormal behavior detection have brought a great challenge . For more than two technologies proposed an effective detection method : (1) the method with MILBoost pedestrian detection methods based on the gradient direction histogram gradient orientation histogram AdaBoost algorithm , MILBoost - algorithm , the image in pedestrian the quick retrieval . The gradient direction histogram method to extract the characteristics of the object of detection , first characterized in the initial screening by AdaBoost , and then use the combined training MILBoost characterized screened final discriminant desired classification . By a large number of experiments comparing we can see : Add MILBoost pedestrian detection algorithm for the multi- gesture , multi-scale pedestrian good detection results . (2) The method based on optical flow monitor and connectivity described temporal behavior anomaly detection method based on the optical flow method , the monitoring of the scene is divided into a plurality of regions , each region has a monitor may be the motion of the optical flow of the abnormal alarm, according to the occurrence of abnormal behavior is a continuous process , there must be a certain amount of time and space connectivity features , in order to determine whether the occurrence of abnormal behavior . So that we effectively reduce the false detection due to the individual the noise disturbance of optical flow , and improve the efficiency of detection . This paper will use the the AdaBoost algorithm , MILBoost algorithm , gradient direction histogram features and Lucas-Kanade optical flow method and other methods . Each chapter be methods used were given a basic introduction , and then given in detail the proposed pedestrian detection method to detect the abnormal behavior .

Related Dissertations
More Dissertations