Study On Automatic Measurement Technology Of Human Parameters Based On Video

The mass and real-time body measurement of athletes,dancers,soldiers and aerospace and astronauts can be realized by the automatic measurement technology based on video.In the future research,the video-based human parameter measurement technology will be combined with related technologies such as human kinematics analysis,so as to build an electronic referee system that can automatically score gymnasts’ movements.Video-based automatic measurement technology of human parameters has been widely used in the fields of ergonomics,costume design,virtual fitting and virtual human simulation.A set of actions for measuring human parameters was designed in this paper.By detecting human objects and special body parts such as eyes and hands in videos,the extraction of human feature points and the location of feature regions were realized.Real-time dynamic measurements of 10 human body parameters including typical standing posture,head and hand parameters were completed.The major research contents of this paper include:(1)In order to solve the problems of poor real-time performance and inability to measure dynamic human parameters according to traditional human parameter measurement scheme based on static image,the shooting method of human parameter measurement videos and a set of auxiliary measurement actions of standing posture,head and hand measurement items were proposed.(2)An improved HOG+SVM pedestrian detection algorithm based on ViBe was studied.On the basis of foreground extraction,the HOG feature based on integral histogram was used to detect the head region of pedestrian.By extracting the head and foot points of pedestrians,the dynamic measurement of human walking height was realized.Experimental results demonstrate that this method is simple and valuable.(3)An eye detection algorithm based on Haar classifier and Kalman was studied.Eye detection was realized on the basis of face detection,and the detected eyes were tracked by Kalman algorithm.Automatic measurements of interpupillary distance and eye height were realized by locating the position of human eyes in videos.(4)An adaptive hand detection algorithm based on HOG-LBP feature was studied.HOG feature and LBP feature were fused according to the confidence.The new feature obtained by fusion was used to train SVM classifier,so as to increase the accuracy of palm and fist detection.By positioning the hand position under specific actions in videos,the real-time measurements of seven human parameters,such as the middle fingertip height overhead and forearm-hand length were realized.Through the comparative analysis of the errors between the video measurement results and the real measurement values of various human parameters,it is proved that the automatic measurement technology of human parameters has the advantages of strong operability,high accuracy and good real-time performance.