Research And Implementation Of Video Based Abnormal Behavior Detection Algorithm In Elevator

With the rapid development of urbanization,a large number of high-rise buildings which relying on elevator system emerge,how to improve the safety and efficiency of the elevator has become an urgent need.This thesis analyzes the existing safety problems of elevator,including the personnel overload,fighting,placing dangerous goods and other passengers' abnormal states.Without human intervention,using surveillance video information to analyze the elevator passengers status,and alarming if abnormal states are detected has good engineering application value.This thesis researches the video-based elevator passengers counting algorithm,elevator personnel abnormal behavior detection algorithm and elevator abandoned detection algorithm.This thesis adopts VIBE background model algorithm to detect moving objects.Adaboost and SVM are cascaded,and hough circle transformation is assisted to identify the human head.The MedianFlow algorithm is used to track and count passengers to implement passengers counting algorithm.Using motion estimation to extract motion information,this thesis improves the traditional three step search method in search strategy and matching function.Calculate the entropy of the moving area,and then judge the abnormal behavior to implement personnel abnormal behavior detection algorithm.This thesis adopts dual VIBE background model,morphological filtering and connected component analysis to extract the target foreground.After judging the object is abandoned,alarm immediately to implement abandoned detection algorithm.Finally through the experiment of abnormal behavior detection algorithm in elevator,this thesis implements real-time and accurate passengers counting,passengers'abnormal behavior detection and abandoned detection.