The Segmentation And Recognition Model Of Middle-aged Male With Convex Belly Based On Non-contact Anthropometric Measurement

With the rapid development of the economy,people are more likely to pursue a kind of personalized demand at nowadays,and the generation and continuous updating of non-contact measurement provides a way to meet people’s individual demand.In the era of industrialization of mass production,most clothing manufactures produce standardized and"only one face" clothes to make its products cover the most of the people and obtain the maximum profit at the minimum cost.These products meet the clothing fit needs of most people,but ignore a small number of special body types.At the same time,the development of the clothing industry become intelligent,automated and digital what has reawakened people’s attention to special body types.The changes in people’s diet have brought about changes in the national body type,and the number of special body types has gradually increased.The deformation and shape of the body are irreversible follow the increase of people’s age.Especially middle-aged male whose waist and belly present different degrees of bulging,and these people form a large crowd of special body type with convex belly.How to identify the complex and diverse special body types with the help of 3D anthropometric technology,how to subdivide and recognize special body types automatically,how to meet the dress fit and comfort demands of special people is a series of key problems that need to be solved on the way to intelligent and automatic tailored clothing in the field of clothing.This subject is based on the non-contact measurement of the bulging bell of middle-aged male body segmentation and recognition model.The purpose is to extract the characteristic indexes by analyzing the morphological characteristics of the section of the corpulent abdomen body to characterize each part of the corpulent abdomen body.The range of each part of the special body was divided so as to form the subdivision of the convex belly body through the statistical analysis of the characteristic values.Then the BP neural network is used to construct the recognition of convex belly.So as to explore a way to subdivide and automatically identify the sprcial body type of the convex belly.The method provides a theoretical basis for the recognition of special body types,personalized tailoring,and virtual fittings for 3D anthropometric techniques.The main research contents of this topic are:(1)Experimental design.Using the Size Stream 3D body scanner to measure 420 middle-aged male and 404 valid samples were obtained at last.Then,five characteristic indicators such as height,chest circumference,waist circumference,hip circumference and body weight were selected for k-means cluster analysis.The 404 middle-aged male were divided into five categories:165/90B,170/82A,170/96C,175/104C,and 175/90A.The C-type obese body type can account for 33%of the total sample which provides certain feasibility for our further research.(2)Waist curve feature value extraction.The imageware reverse engineering software was used to extract the waist section curve of the human point cloud model,and obtain the value of transverse radius and perform cluster analysis to determine the range of the convex belly body.Among them,the group with a waist-to-waist diameter ratio of 0.77～0.82 was defined as a slightly convex belly body and it accounted for 23%of the total sample,and the group with a waist-to-waist diameter ratio of 0.83～0.90 was defined as a severe convex belly body and it accounted for 7%of the total samples.(3)Obtain the segmentation index of the convex belly body.Point out the other special parts that are easy to occur with the convex belly body,such as the back,waist,buttocks and so on.Then,the surface indexes such as the back angle,the dorsal angle,the hip convex angle and the lumbar convex angle a and b which characterizes the convex were extracted to refine and classify the special body of convex belly.Imageware reverse engineering software was used to extract the four characteristic sections of the sample,including the central sagittal plane,the sagittal plane of the cross dorsal convex point,the sagittal plane of the cross buttock convex point,and the cross section of the cross lumber convex point.AutoCAD was used to fit these cross-sectional figures and extract feature point coordinates.Finally,a total of 535 feature angles of 107 convex belly body samples were calculated.(4)Statistical analysis of feature angles.Statistical analysis and correlation analysis were carried out on the five characteristic angles,and four characteristic angles such as the back angle,the backside angle,the hip convex angle and the lumbar angle a were selected for final refinement classification.Regression analysis was carried out between the four characteristic angles and the easily measurable human body parts,and the regression equation was obtained.(5)Refined classification of convex belly.Four convex angle samples were used to classify the convex belly samples,and the three sagittal images of the four clustered cluster center samples were drawn.Finally,combined with the classification results and image morphology,the convex belly body was divided into four categories:the bow-backed,straight-waist and flat-buttock mild convex belly body,the straight-back,straight-waist and flat-buttock mild convex belly body,the normal-back,normal-waist and normal-hip mild convex belly body,the normal-back,normal-waist and normal-hip moderate convex belly body.The four categories accounted for 12.15%,24.30%,40.19%,and 23.36%of the total sample of the convex body respectively.(6)Automatic identification model construction phase.The basic principle of artificial neural network is introduced.The BP neural network is used to construct the recognition model of the middle-aged male subdivision type of convex belly.The overall recognition rate of the recognition model is 92.6%,in which the recognition rate of category 1 is 100%,the recognition rate of category 2 is 100%,the recognition rate of category 3 is 90.9%,and the recognition rate of category 4 is 85.7%.The overall recognition rate is high and has certain application value.Through the above research,the method of subdivision and recognition of the special body types of the convex belly is constructed.The method can provide a research path for the subdivision and identification model construction of other special bodies and provide a method for the non-contact measuring equipment to automatically identify the special body type.It also can lay the foundation for the development of personalized customization.