基于YOLOv8的居家环境跌倒行为识别

Fall Behavior Recognition in Home Environment Based on YOLOv8

对于现有跌倒行为识别算法在复杂的居家环境条件下,出现算法精度低、实时性差等问题,文章提出一种基于YOLOv8的居家环境跌倒行为识别方法。该方法通过网络摄像头获取视频图像,使用基于YOLOv8的目标检测算法识别监控视频中每一帧画面的人体与跌倒特征,再结合时序状态特征处理,设定规则判别跌倒行为,并进行跌倒预警。实验证明,改进的方法精确率达94.9%,召回率达95.7%,FPS为40,算法识别准确率高、实时性强,为跌倒行为识别提供了一种简单而有效的方法。

For the algorithm problems of low accuracy and poor real-time performance of existing fall behavior recognition algorithms in the complex home environment conditions,this paper proposes a fall behavior recognition method in home environment based on YOLOv8.This method obtains video images from webcams,uses object detection algorithm based on YOLOv8 to identify the human body and fall features in each frame of surveillance video,and then combines the processing of sequential state features to set rules to identify fall behaviors and conduct fall warning.The experimental results show that the precision rate of the improved method is 94.9%,the recall rate is 95.7%,and the FPS is 40.The algorithm has high recognition accuracy and strong real-time performance,which provides a simple and effective method for fall behavior recognition.