Research On Automatic Recognition Of Male Body

Body shape analysis is a key link in the production of MTM. Accurate body shape analysis is an important guarantee to meet the demand of consumer. It is a prerequisite for personalized clothing customization to start from the customer’s own body, to establish the perfect body shape analysis system, to improve customer’s satisfaction for custom clothing. Combined with silhouette of men’suits and front contour feature of male body, this topic determine the method by using profile of positive male body type to differentiate human body, and to distinguish the male body side and the local feature size. Realized the new body shape classification from the whole to the part and from the details. From analysis of data collected from human data, index classification, verification to testing validation, this paper did the following works.First, determine the method and index by analysis and processing data of human body, so as to realize the division of human body, this part include the following three aspects:(1) From questionnaire of men’suit and male body size analysis to determine classification index of male body shape:proportion between shoulder, waist and hip width to differentiate X, H, V three kinds of profile as the front of body shape. Using back angle to determine chest and back shape, ventral lobe said the lower abdomen shape, rump angle determine hip shape, shoulder angle determine the shoulder shape.(2) Using [TC]23D scanning measurement system collected311male body data between the ages of25-55in Hunan area. Matlab programming to recognise three dimensional human, pre-processing point cloud data through the point cloud data analysis, point cloud smoothing noise and reconstruction. Recognise characteristic points of3D human body after pretreatment, then extract feature size. Analysis the extracted data, got290samples that can be used after excluding21singular value samples obtained. Correlation analysis was carried out on the index size to determine the index relationship.(3) By division method to classify human torso length, and the method of fuzzy clustering analysis the ratio between shoulder and waist and hip width to determine the sample size and the final cluster centers. Using the method of cluster analysis, and combining with the angle index normal distribution to determine the different morphological characteristics of shoulder and side angle corresponding to value. Got12main kinds of human body under the shoulder and side form of290samples.Subsequently, establish the automatic identification system. On the basis of BP neural network to establish positive pattern recognition model, then combined with the interval comprehensive of shoulder and morphology index by cluster analysis to discriminant shape Established by the Matlab programming, the automatic recognition system used to achieve the following functions:automatic preprocessing three-dimensional human body model, extract size operation; automatic identification, discriminant morphological characteristics of each part of the human body; automatic storage feature size and the judgment result.Finally, the shape classification method is applied to the personalized production of suit and validate the method. Selected three validation samples, and got their body size through the automatic recognition system. On the basis of the positive form to recommend suit styles, then based on the shoulder, back, abdomen and buttock type to modify pattern. Made suits according to the modified model, and the subjects and expert scoring evaluation respectively with the validation sample fit the suits.It is the innovation of this research that proposed a new method to analysis human body, which use a positive profile of human body and convex and concave feature of human side and shoulder to divide the human body under the global and local train thought. The somatotype method was validated in enterprise’s practice.