基于轮廓跟踪的摔倒检测算法  

Fall Detection Algorithm Based on Contour Tracing

作　　者：靳海伟[1] 彭力[1] 卢晓龙[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《江南大学学报（自然科学版）》2015年第2期172-177,共6页Joural of Jiangnan University (Natural Science Edition)　

基　　金：江苏省产学研联合创新项目(BY2014023-36;BY2014023-25)

摘　　要：针对传统摔倒检测算法误判率高和算法复杂等缺点,提出一种基于轮廓跟踪的摔倒检测算法。该算法主要通过背景减除法提取监控视频中的运动目标老人,运用中值滤波和形态学处理获取完整的老人身体形状,根据轮廓跟踪算法生成边界链码,然后利用实际边界链码与摔倒模板链码的欧氏空间距离判断老人是否摔倒。实验结果表明,该算法计算复杂度低,准确率得到明显提高。In view of the shortcomings of high misdiagnosis rates and the complexity of traditional fall detection algorithms, a fall detection algorithm based on the contour tracing is proposed. The algorithm extracts the moving target in the surveillance video by background subtraction,median filtering and morphological processing are used to get a complete body shape. Generating the boundary chain code with contour tracking algorithm, then the eider fall or not is determined by Euclidean space distance between actual boundary chain code and the fall template chain code. Experiments show that the algorithm reduces computational complexity and improves accuracy rate significantly.

关 键 词：摔倒检测算法 背景减除法 轮廓跟踪 边界链码 欧氏空间距离 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]