Classification And Recognition Model Of The Hunchback Young Male Body Based On Shoulder Section Curve

With the development and application of non-contact body measurement technology,the fashion development gradually towards to the direction of digitization and automation.High quality and personalized needs of clothing consumption spread gradually,but people’s diet and lifestyle changes with every passing day,led to the overall national body shape changing,so the present clothing standard in garment industry can’t meet fully the consumer on the clothing comfort,fit and personality requirements,especially for the crowd with special body size.At present,the definition of hunchback is a vague concept,mostly are divided into hunchback body and scapula body based on their special body characteristics,haven’t indicated standard and normatively the degree of arched back,detailed classification and identification methods.The shoulder section curve shape of hunchback body is closely related to clothing shoulder structure,which is lack of the distinction between shoulder and cross section curve shape characteristics.Moreover,there is little research on shoulder section curve,which really affected the development and popularization of personalized customization.While how to quickly and accurately identify the human body size and special parts based on the non-contact body measurement technology,how to improve the size breakdown and automatic identification technology,how to quickly and accurately acquire the shape changes of hunchback body’s special parts,how to meet the special consumers’ personalized needs for clothing comfort and fit,which need to be further explored and solved on the way to realize personalized customization in the garment industry.The purpose of this paper is to identify more detailed shape features of the hunchback’s should section.This study constructs basic shoulder-section curve mathematical of the section curve,the hunchback young men body as the research object,and extract the shoulder section curve morphological parameters to represent the curve shape and feature,which provide a new method for hunchback body.At the same time,building the automatic identification model based on probabilistic neural network to identify the hunchback quickly and meticulously,which provides technical support for personalized customization,body size segmentation and identification technology and virtual fitting.The thesis mainly includes the following contents:(1)Non-contact anthropometric and body cluster analysis: using the Size Stream non-contact body measurement instrument subject to measure 320 young men according to standard measurement condition,exporting the necessary body data from the measurement system to preprocess and obtain 302 sets of valid body data.Using K-Means clustering method to classify the male upper body by selecting height,chest circumference,waist,shoulder circumference and the ratio of chest circumference and height with making multiple correlation analysis as size classification basis,got 170/84 A,175/92 A,180/88 A as shape clustering centers.(2)the hunchback selection: the parameter named shoulder arch which is the ratio of front shoulder width and shoulder circumference was established to study shoulder horizontal camber or shoulder blade projection;by the way of AGNES clustering analyzing the shoulder arch was analyzed and different shoulder models were acquired,and selecting small shoulder arch as hunchback body.According to the value of shoulder arch,different degrees of the hunchback were defined,the value of mild hunchback was between 0.721 and 0.813,the value of severe hunchback was less than 0.721.(3)buildingmathematical model of shoulder section curve: the hunchback shoulder section curve was extracted and processed by smoothing,rotating and symmetry.Using AutoCAD software to obtain equidistantly 110 fitting points,and using Matlab to build curve mathematical model.By analyzing and comparing on imitative effect,error and goodness,cubic and quinticpolynomial were used to fit back shoulder curve and front shoulder curve,and the goodness more than 0.98,so the mathematical model had high accuracy.(4)extracting characteristic parameters of curve: the radius of curvature of all fitting points were calculated based on curve mathematical model,and got midpoint,shoulder blade,shoulder point,clavicle depression and clavicle protrusion point as curve’s characteristic points.So characteristic parameters of curve were the radius of curvature of characteristic points and the ratio of breadth and length,which had correlation and regressionanalysis with human body data measured easily so that characteristic parameters of curve could be obtained quickly and easily.(5)detailed classification of the hunchback: characteristic parameters as classification index,using pseudo-F statistics to determine the best classification number,based on the GB standard,making use of K-Means method to divide 170/84 A,175/92 A,180/88 A into 4,3 and 3 categories,analyzing their morphology and corresponding clothing pattern.(6)constructing automatic recognition model of the hunchback: according to the detailed classification results,constructed a automatic recognition model of the hunchback body based on the probabilistic neural network by Matlab software programming language.The model was trained and tested,its overall recognition rate was 91%,the recognition rate was high,and it had certain accuracy and feasibility.