摘要
随着人口老龄化问题日益严重,老年人的安全问题变得愈加重要,而对老年人安全问题威胁最大的便是老年人的跌倒问题,因此文中提出了一种基于关节点提取以及SVM分类器的老年人跌倒检测算法。首先通过改进的yolo算法检测出视频帧图像中的人所在的位置,然后将单个人的图像送入openpose算法中获得这个人的关节信息。之后再通过SVM分类器对获得到的关节点信息进行分类,以得到这个人所处的状态(此处将人可能处于的状态分为4类——正常状态、跌倒状态、平躺状态以及其他状态)。对于整段视频便得到一个状态序列集,之后对这个状态序列集进行分析便能够检测出视频中是否有跌倒事件发生。经实验对比,该算法对于单人存在的场景有着98%以上的准确率,并且对于多人存在的场景具有一定的鲁棒性。
With the increasingly serious aging of the population,the safety of the elderly becomes more and more important,and the biggest threat to the safety of the elderly is the fall of the elderly. Therefore,we propose a fall detection algorithm for the elderly based on joint point extraction and SVM classifier. First of all,the location of the person in video frame images is detected by the improved yolo algorithm,and then the image of an individual is sent to openpose algorithm to obtain the joint information of the person. Then,the obtained joint point information will be classified by the SVM classifier to get the state of the person (There are four possible states of being – normal,falling,lying flat,and others). After that,we will get a state sequence set which is analyzed to detect whether there is a fall event in the video. Through experimental comparison,the proposed algorithm has an accuracy rate of more than 98% for scenes with single presence and robustness for scenes with multiple presence.
作者
袁鹏泰
刘宁钟
YUAN Peng-tai;LIU Ning-zhong(School of Computer Science and Technology,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)
出处
《计算机技术与发展》
2019年第9期200-204,151,共6页
Computer Technology and Development
基金
国家自然科学基金(61375021)
南京航空航天大学研究生创新基地(实验室)开放基金(kfjj20171608)
