Design And Implementation Of Human Fall Detection System Based On Machine Vision

With the continuous development of embedded system technology and machine vision technology,the human body fall detection system is developing in the direction of intelligence and practicality.Studies have shown that fall in the elderly can easily lead to severe disability or even death,and more than half of the fall occur in indoor environments.In view of the hidden dangers of the elderly living alone,if the elderly can be detected in time and the alarm information can be sent to the guardian in time,the injuries to the elderly due to the fall will be greatly reduced.For this purpose,this thesis uses the embedded development board EAIDK-310 and Open CV library as the hardware and software foundation respectively,combined with image processing technology,designed and implemented a human fall detection system suitable for indoor environment,realized intelligent detection of fall and timely call the first aid.Our system does not need to be worn by the human body and can monitor the activities of the elderly in real time,thereby reducing the injuries caused to the elderly due to fall.The research work of this thesis is summarized as follows:(1)Analysis of commonly used moving target detection methods,such as background difference method,frame difference method and optical flow method;for system design requirements,based Gaussian Mixed Model to achieve moving target detection,and anti-interference processing is added to improve moving target detection.The method is verified by experiments to obtain a clear and complete foreground outline.(2)Design a fall detection method based on multi-parameter comprehensive evaluation.Analyze various behavior characteristics of human body,use three contour features of aspect ratio,centroid height and tilt angle to represent human posture,and propose a fall detection algorithm based on multi-parameter comprehensive evaluation.Based on the experimental analysis of the changes of various contour features in different scenes,the threshold is set by the contour feature threshold selection algorithm,so that the fall detection can achieve the best detection effect.(3)The embedded detection board EAIDK-310 is used as the hardware platform,and the Open CV library is used as the software development environment to design and implement the fall detection system.This system can effectively distinguish between the fall behavior and the activities of daily living.The image of about 10 frames can achieve the purpose of detecting falls in real time.Finally,the performance of the system is analyzed through multiple sets of comparative experiments.The experimental results verify that the system can achieve the design purpose,detect human falls in time,and have a high detection accuracy.