Design and Development of Posture Detection System Based on YOLO - OpenPose
Yujie Jiang, Rongzhi Hang, Yanhao Wu, Wěipéng Huáng, Xiaoping Pan, Zhi Tao · 2023
With the development of machine vision and multimedia technology, posture detection and other related algorithms have been widely used in the field of human posture recognition. Traditional video surveillance methods have the disadvantages of slow detection speed, low accuracy, interference from occlusion, and poor real-time performance. This paper proposes a posture detection system based on YOLO-OpenPose. Based on the YOLO network, a corresponding human recognition dataset is created for complex scenes. OpenPose is used to detect human key points, and Kalman filtering multiobject tracking method is applied to predict the target state of human objects in occluded areas. Real-time detection of human postures (standing, sitting, walking, falling, and lying down) is achieved with corresponding alarms to ensure the timely detection and processing of emergencies.