Study On Human Motion Measurement System In Monocluar Video

Video-based human motion measurement technology is a new non-contact measurement method,widely used in criminal investigation,intelligent medical,intelligent transportation system and ergonomics and other related fields.In this paper,a human motion measurement system is designed for monocular video.Through the background modeling,detection,tracking and feature extraction the moving pedestrian in the video,the real-time height and trajectory of the pedestrian,speed,stride and step frequency and other biological characteristics and motion parameters can be measured.Through the analysis and comparison of the experimental simulation data and the measured data,it shows that the measurement system is accurate and reliable,and the operability and real-time are better.The main contents of this paper are shown are as follows.(1)A video measurement algorithm is discussed to determine the actual distance between parallel planes under uncalibrated video conditions.A three-dimensional linear model of human body is established to determine the human feature points and characteristic lines.(2)Based on the background modeling and machine learning,an improved HOG + SVM detection algorithm based on background modeling is proposed.On the basis of background modeling of VIBE,integral histogram is used to extract HOG feature of pedestrians' head.The experimental results show that the algorithm is accurate and fast.(3)Combining the Meanshift tracking algorithm and Kalman filter-based tracking algorithm,a new moving pedestrian target tracking algorithm is proposed.The experimental results show that the algorithm can solve the problems of pedestrian target overlap and background interference,and can meet the real-time measurement requirements with high accuracy and strong robustness.(4)The monocular video motion measurement system of the human body is designed.The simulation of calibrated video measurement is carried out and the actual experiment is carried out for different scenes.The experimental results are compared and analyzed.Results show that the measurement system is highly precise and robust.