Model Catwalk Movement Research Based On Motion Capture

In recent years,human motion recognition has caused widespread concern.Visionbased motion capture systems use a camera as a sensor,which is free from scene constraints when capturing human movements and has obvious advantages compared with traditional solutions.In the field of fashion performance,most of the current evaluations for catwalks use manual evaluation methods.By capturing and analyzing the target motion data through computer vision technology,it can realize the automatic evaluation of the model's walking posture,and produce more comprehensive quantitative indicators,which are not affected by subjective factors.The development of a model of posture-evaluation based on motion capture is of great significance in the teaching and training of assisted models.This paper systematically introduces the development and current research status of motion analysis technology based on computer vision.Then,for the specific scene of the model catwalk,a model walking posture evaluation method based on gait segmentation and template comparison is proposed.The specific work is as follows:1.The bottom-up human skeleton model detection method is used to extract the joint information of the moving target in the image,and solve the defect that all the joint points in the target are lost due to the loss of the human target in the previous topdown method.2.Performe gait segmentation on the formed sequence of actions.The principal component analysis is used to compress the angular features of the joints with obvious changes in the cycle,and the one-dimensional features formed can exhibit more stable periodic changes.Then through the operations of dimensionality reduction,filtering,peak finding,etc.,the key frame positions in the gait cycle are found in the formed features to realize the segmentation of the action sequence.3.Extract a gait sequence with a fixed period length and comparing it with a standard template sequence,calculate the distance between the two as the basis for evaluating the standard of walking posture.Since the joint point data used in this paper is two-dimensional plane data instead of three-dimensional data,for the two known starting and ending sequences,this paper introduces Procrustes analysis on the traditional DTW algorithm.The joint points on the group plane realize the alignment of the joint point coordinates by affine transformation and reduce the error caused by the camera shooting angle as well.This paper implements the MFC-based model catwalk evaluation system.The index results of the system include three joint point indicators and three detailed parameter indicators,which can evaluate the standard degree of the model's walking posture in many aspects.The experimental results show that the system has a good ability to evaluate the posture.