Research On Detection Method Of Human Behavior And Expression Abnormality Based On Video

Video anomaly detection is a hotspot in the field of image processing and artificial intelligence.The intelligent detection of abnormal behavior and abnormal expression in surveillance video can not only reduce the cost of manual detection,but also deal with emergencies in time.And it plays a important role in the field of security and defense.Therefore,the detection of abnormal behavior and expression has attracted more and more attention.This paper focuses on video-based detection of human behavior and facial expressions.The main objects of abnormal detection are the elderly and young children.The anomaly is mainly divided into two parts,one is the behavior anomaly detection of the detected object,the other is the expression abnormal detection of the detected object.In the detection of fall and squat behavior,the first step is to extract the moving target and to deal with the moving object morphologically.The hybrid Gao Si model and background differential method are used to detect moving targets.Then,in the research of behavior anomaly detection,the detected object is identified and tracked by the external rectangle,and the center of gravity of the detected object is represented by the gray centroid of the pixel point in the outer rectangle.Two characteristic operators,the variation rate of barycenter height and the ratio of height to width of the outer rectangle,are defined,and the anomalies of moving targets are judged by observing the changes of the two operators.In the detection of crying,the face is first detected based on skin color and AdaBoost method,and the feature extraction algorithm of LBP and LPQ is used to extract the facial expression features in the video sequence.Then the histogram sequence of the video sequence expression feature image is extracted.Finally,weightng the extracted two histogram sequences and generating a histogram sequence in series.Compared with the LBP and Gabor feature extraction algorithms,the feature extraction algorithm based on LBP and LPQ has a higher recognition rate for crying anomaly detection.Finally,experimental verifications of three abnormal behaviors of falling,squatting,and crying were performed on multiple data sets collected by this article.The results show that the algorithm used in this paper has a good detection effect.