Research And Implementation Of Fall Detection Based On Posture Analysis

With the aggravation of the population aging problem,the research on fall detection is of great significance.The video-based fall detection method has become a research hotspot in the field of fall detection due to the characteristics of convenience and efficiency.In this paper,by analyzing the characteristics of human body posture,the method of falling detection based on human body posture features is studied.In terms of moving objects extraction,this paper analyzes the Vi Be algorithm and uses the improved Vi Be algorithm to extract foreground objects.The improved algorithm can effectively overcome the defects of "ghost" and "void" phenomena in the original Vi Be algorithm.The algorithm constructs a virtual background by averaging multi-frame images,and then performs background modeling.It replaces the fixed threshold in the original Vi Be algorithm with an adaptive radius threshold,and then removes the detected scintillation pixels and fills the foreground region.Compared with the original Vi Be algorithm,the improved algorithm can effectively eliminate the phenomenon of "ghost" and extract the foreground objects more completely.The characteristics of the human body posture are analyzed,and the four characteristics of the body aspect ratio,height change rate,direction angle and motion history image are selected to judge the fall.Firstly,the feature of aspect ratio is used to eliminate the normal walking situation,which greatly reduces the complexity of the algorithm.Then the human body height change rate is used for further detection,which can avoid the occurrence of false detection caused by accidental bumping but not falling.Then,the posture of the human body is further judged by the size of the direction angle.Finally,through the relevant characteristics of the motion history image,the global movement direction of the human body and the speed of the movement are judged,and then it is judged whether the fall situation occurs.Based on the algorithm implementation,this paper designs a visual display interface based on MFC,so that the experimental results are displayed on the same interface,which has good stability and scalability.By selecting buttons to operate,human-computer interaction is convenient,it is easy to operate and manage.The experimental results show that the accuracy of the fall detection method proposed in this paper is 92.5%,it takes 39 ms per frame and has good real-time performance,it can be applied to smart home and has certain practical value.