基于深度数据分析的室内老人摔倒检测方法

摘要

人口老龄化是当前面临的重要社会问题,摔倒检测是老年护理和生活中的一个重要课题.针对基于穿戴传感器、环境传感器和计算机视觉的摔倒检测系统具有高入侵性、低精度和鲁棒性差等缺点,提出了一种基于深度数据分析的室内独居老人摔倒检测方法.微软Kinect传感器可有效、廉价地获得场景的三维信息,该方法基于Kinect传感器捕捉的深度数据进行摔倒检测分析.首先对背景帧深度数据和目标帧深度数据使用滤波方法进行预处理,并采用高斯模型对运动目标进行检测;然后通过水平和垂直投影直方图将深度图转换为视差图以得到地面信息,并用最小二乘法估计地面方程的参数;最后分析人体对象的动态信息,针对室内独居老人摔倒检测,假设室内只有一个运动个体,对于目标帧深度图像,计算人体重心与地面之间的距离.当人体重心到地面的距离低于阈值时,判定摔倒行为发生.实验结果表明,该方法可以有效地检测摔倒事件.%Aging is a serious problem now.Fall detection is an important subject in aged life care.The fall detection system based on wearable sensor,environmental sensor and computer vision has invasion,low precision and poor robustness and other shortcomings.For this we present a new fall detection method of elderly people living alone in a room environment based on the depth data analysis.Mi-crosoft Kinect sensor can get the three-dimensional information of the scene effectively and cheaply,and the method can carry on the fall detection analysis based on the depth data captured by Kinect sensor.Firstly,the depth data of both background frame and target frame is preprocessed by the filter method,and the moving object is detected by Gaussian model.And then the depth images are converted to dis-parity map through the horizontal and vertical projection histogram to get the ground information,and the least squares method is used to estimate the ground equation.Finally,the dynamic information of the human body is analyzed.It is assumed that there is only one moving individual in the room for the fall detection of elderly people living alone.Distances between the center of gravity and the ground are cal-culated through the target frame of depth image.When the distances from the centroids of the human body to the ground are lower than some thresholds,fall incident will be detected.Experiments show that the proposed method can detect fall incidents effectively.