Dissertation > Industrial Technology > Automation technology,computer technology > Automation technology and equipment > Automation systems > Data processing, data processing system > Centralized testing and roving detection system

Intelligent Video Analysis Technology for Elevator Cage Abnormal Behavior Detection Based on Computer Vision

Author LuHaiFeng
Tutor TangYiPing
School Zhejiang University of Technology
Course Applied Computer Technology
Keywords Computer Vision Abnormal behavior detection The number of judgments Clustering algorithm Contour Extraction The Hidden Markov husband model Eigenvectors
CLC TP274.4
Type Master's thesis
Year 2009
Downloads 97
Quotes 0
Download Dissertation

Increasing number of security issues need to be more concerned about the violence against the elevator car. Traditional video surveillance can not automatically detect abnormal behavior, also need people to participate in its efficiency and accuracy is not satisfactory. Intelligent video surveillance technology as an inevitable development trend of anomaly detection in the elevator car, able to overcome the shortcomings of the traditional video surveillance to identify abnormal behavior. This paper studies the key technology of digital image processing, pattern recognition and computer vision in intelligent video surveillance, detection of abnormal behavior of an elevator car. Extraction algorithm based on the the background subtraction human prospects of the four first study to extract the prospects of the human body, and then judged by the number of regional pixel statistics based on the prospect of communicating method to obtain information on the number of elevator car finally taken for single and multiplayer different abnormal behavior detection. For many people, the main detection is similar to fighting this abnormal behavior. By the acquisition of the video image within the elevator car, and calculating the change in the number of human foreground pixel, foreground external rectangular aspect changes and prospects of a circumscribed rectangle of center variation of these three related feature, and form a three-dimensional body-movement feature vector. In this paper, three clustering methods to obtain three-dimensional feature vector and feature vector data using clustering algorithm, observe the sequence of symbols. Observed symbol sequence using the obtained body's normal behavior patterns on the elevator car to create a hidden Markov model based on the hidden Markov model with the normal behavior of the comparison to identify the abnormal behavior of the elevator car multiplayer. For a single, major detection is similar to the sudden illness and fell to the ground long stationary abnormal behavior. Firstly, by studying the human body contour tracking method based on binary image to obtain the initial outline of the human body; then study the Snake, and closer to the body shape of the body contour obtained by this method; Finally Hausdorff algorithm for consecutive two images human contour matching degrees calculated by the the contour matching degree to calculate the period of time within the human body before and after the frame, to determine whether the human body is in an abnormal state of the long stationary. The developed smart detection system based on the abnormal behavior of the elevator car in the Java language, described in detail the implementation of each module, and related experiments in a simulated environment. The experimental results show that this method can effectively detect people fighting in the elevator car and single people for a long time still abnormal behavior.

Related Dissertations
More Dissertations