Research On Indoor Fall Behavior Detection Based On Video Monitoring

With the aging of the society,accidental falls have become a serious threat to the health of the elderly.It is necessary to detect and alarm the fall behavior of the elderly accurately,so as to reduce the injury of the fall to the elderly,which has high research value and practical significance.Compared with other methods,the fall behavior detection method based on video monitoring has many advantages.Therefore,this paper studies the indoor fall behavior detection method based on video monitoring.First of all,based on the analysis of the advantages and disadvantages of various moving object extraction algorithms,combined with the requirements of the project for the extraction effect of moving objects and the characteristics of indoor scenes,this paper uses the background difference method based on the mixed Gaussian model(GMM)to extract moving objects,and uses morphological methods such as expansion and corrosion to further process the extracted moving objects to fill in Finally,the whole moving target is extracted accurately.Secondly,in order to achieve the accurate tracking of the human head in the video,this paper designs an automatic head tracking algorithm based on particle filter,and proposes that the human head position information extracted from the moving target is used for the initialization of the tracking position,and then realizes the accurate tracking of the human head in the video.Then,in order to obtain the effective fall detection features,based on the analysis of the common features of fall detection,this paper adopts the feature extraction method which combines the shape feature(aspect ratio of human body)and the head motion feature.In order to eliminate the influence of the distance between the human body and the camera caused by the relative motion between the human body and the camera in the video on the feature extraction results,this paper proposes to use the relative values of the head motion speed and head height as the motion features of this paper,and defines and extracts them respectively,and further analyzes each feature The feasibility of the sign as a fall detection feature in this paper.Finally,the feature threshold value is determined through the analysis of the extracted fall features.The initial verification and evaluation of the fall detection algorithm is carried out on the collected fall behavior data set by using the feature threshold method.Seven indoor activity video data including the fall behavior are collected by using the camera.The fall detection algorithm in this paper has The effectiveness was further verified.Experiments show that the algorithm has high accuracy and reliability in the data set and real environment.