Research Of Indoor Abnormal Behavior Detection For Lonely Elder Based On Video

With the rise of intelligent home,safe city and intelligent city construction,accelerate the demand for intelligent video surveillance,and with the development of our society and the aging of the population increase,more and more elderly at home alone,the guardians of these groups which has become a problem that cannot be neglected.Therefore,it is of great significance to carry out intelligent detection and research on the behavior of the elderly in the room.The research direction has great prospects and social benefits.In this paper,after analyzing the common behaviors of the elderly living alone,this paper presents and realizes a video-based indoor abnormal behavior detection system,which is used to analyze the daily behavior of the elderly at home alone,discover abnormal behaviors and reduce family losses.The system mainly includes the indoor detection of human body,tracking the target and extracting the trajectory and the abnormal behavior of the elderly for analysis and classification of these parts.Firstly,an improved hybrid Gaussian background modeling algorithm is used to extract the video foreground part,and the learning rate of each pixel is dynamically updated,which can accelerate the adaptability to the changed background.Then,a DPM-based accurate target judgment method is proposed.The foreground part of the background model is sent to the DPM for detection,and the human body part can be obtained more accurately,to a certain extent,at the same time,you can improve the accuracy of tracking targets.Then,in order to adapt to some requirements of the indoor surveillance scene,the tracking can be stably carried out in the case of deformation and occlusion of the target motion.In this paper,a moving object tracking algorithm based on the perceptual hash is proposed.By using the linear interpolation algorithm,and the difference between each trajectory is reduced.At the same time,the finite state machine is combined to keep the system stable tracking,which lays the foundation for the following feature vector extraction and the accurate classification of the behavior.Finally,the feature points of each frame in the trajectory are extracted,and the feature vectors are composed of the trajectory points in successive frames.These feature vectors are trained off-line using support vector machines,and the behavior classifiers are obtained.Then the behavior classifier is used to classify and judge the indoor behavior of the elderly.On the basis of the video database constructed by the author,the validity of thisalgorithm is verified.The results show that the correct rate of recognition of normal behavior is 91.6%,the accuracy of abnormal behavior recognition is 92.7%,and the overall correct rate is 91.9%.It can be used in indoor home scene,and has important significance to protect the safety of his family.