| In recent years,with the rapid development of the Internet,e-commerce and online consumption.It is a new challenge to obtain the human body size quickly.How to understand the characteristics of consumers through the way of image,etc.,targeted to recommend the style and structure design has become one of the most difficult research of human digital clothing..In this paper,we study the differences in the shape and shape of the body under the two-dimensional contours of the human body,and classify the body shape according to the subjective and objective methods.And the shape and of the human body is described by the Shape Context method.The body type library and the body type model are established to identify the body shape,and the recognition result is analyzed to obtain the classification of the body suitable for this recognition model.The main research contents are as follows:(1)Human measurements and data pre-processingIn order to obtain accurate and effective anthropometric data and image information,this paper has carried on the human body measurement work and the image acquisition experiment to 53 samples to be recognized,and the human body size data of all the samples,the three-dimensional scanning human body model and the experimental In this paper,the human body size data and the human body contour curve are obtained,and the method of extracting the contours of human body in 2D graphics is obtained.(2)Body shape description and recognition based on Shape ContextThrough the pretreatment of 53 human body images and 459 human body 3D models,the contour curves of two-dimensional human body contours were extracted and elliptical Fourier method was used to normalize and obtain the contour curve and transfer body's contour curve into a contour point set form.The Shape Context method is applied to the body shape description for the first time.The contour set of the obtained contour points is described by the logarithmic polar coordinate transformation and the histogram of the shape context,and the body recognition framework is established.The similarity measure is made with the established body type library or the body type template,Identify,and to identify the accuracy of the identification.The classification of the shape of the thoracic and thoracic shape under the contours,pelvic morphology and subjective classification of the human body can be well identified in the recognition model of this paper,which has certain practical significance.(3)Building a human body shape classification systemIn this paper,the K-means clustering method is used to analyze the shape of the human body,and the body shape is classified from the perspective of the contours and the lateral contours.The body profile of the trunk part is divided into Type H,Type X,Type A and Type O.The body shape of the trunk part of the side profile is divided into Type HO,Type d,Type C and Type S.The hip shape is divided into flat hip,round hips and drooping buttocks;the chest shape is divided into high-disc breast,the standard hemisphere breast and low-pituitary breast.The size of the chest has been divided into AA,A,B,C +.The pelvic morphology was divided into pelvic anterior and pelvic normal bodies.To the shape of the difference between the chest and back,invited experts to chest and divided back shape into hump back body,intermediate back body,quite back body.The shape of the pelvis,the shape of the buttocks,the size of the buttocks,the size of the chest cup to establish a certain number of body type library,the side of the outline of the chest shape,chest and back shape to establish a body type template,to complete Standard System for Classification of Human Body.In this paper,the relationship between body size and shape is established,and a variety of classification criteria are established.The classification is based on the Shape Context method.The results show that this method can identify the contours of the trunk contours,the shape of the chest and back,the shape of the buttocks and the pelvic form effectively. |
PDF Full Text Request