论文部分内容阅读
为了迅速获得老人跌倒信息,使其得到及时救助,设计和实现了基于嵌入式视频监控的摔倒检测系统。该系统无需人体佩戴任何设备,即可自动检测老人跌倒信息并报警。对监控到的视频流采用H.264标准进行编码压缩,运用B/S网络结构模式实现远程视频监控。同时,对监控到的视频流进行人体摔倒检测以实现报警。摔倒检测使用改进的高斯混合模型算法对背景进行更新,采用背景减除法来分割运动目标。经实验表明该嵌入式系统可实现远程监控并准确分割出人体目标,对其姿态进行自动判别,出现跌倒时进行报警。该系统满足实时性要求,误报率低,鲁棒性好。
In order to obtain the information of the fall of the elderly rapidly and get timely assistance, a fall detection system based on embedded video surveillance was designed and implemented. The system can automatically detect the elderly fall information and alarm without wearing any equipment. The monitored video stream using H.264 standard encoding and compression, the use of B / S network structure to achieve remote video surveillance. At the same time, the monitored video stream for human fall detection in order to achieve the alarm. Fall detection uses a modified Gaussian mixture model algorithm to update the background, using the background subtraction method to segment the moving target. The experiment shows that the embedded system can achieve remote monitoring and accurate segmentation of the human target, to automatically determine the posture, in the event of a fall alarm. The system meets the real-time requirements, low false alarm rate, good robustness.