A Study On Body Type Classification And Size Recommendation Of Female College Students In Hubei Province Based On Neural Networks

With the development of textile and garment industry becoming richer and people’s living standard improving,consumers pursue more comfortable and personalized clothing style,and the production of clothing enterprises gradually develop in the direction of individual tailoring,in which body type analysis is the premise of tailoring,and recommending the suitable clothing size for consumers accurately and quickly according to the body type is one of the problems to be solved.To this end,this project is aimed at the group of female college students in Hubei Province,based on the national standard to study the human body type and clothing size,and build a neural network-based clothing size recommendation system.Firstly,using the German Anthroscan non-contact 3D body scanner and supporting equipment,3D body scans were performed on female university students aged between 18-25 years old at Wuhan Textile University to obtain body shape control part data and 2D images of the front and side of the human body,and the images were imported into Coreldraw software for processing to obtain body surface contour lines,so as to extract the body surface angle.Secondly,we used Spss software to do principal component analysis on body surface angle to obtain five angle indicators for body surface morphology classification,namely,dorsal entry angle,shoulder slope angle,chest convexity angle,lateral body angle and lower hip angle;and used Kmedoids-GMM clustering method to classify body surface morphology into four categories,combined with two classification indicators of chest-waist difference and BMI value to finally obtain a total of 18 subdivided body types.Then Matlab software was used to construct a body type recognition model based on radial basis neural network classification algorithm,and compared with two methods of probabilistic neural network and learning vector quantized neural network,and finally the recognition accuracy of the classification algorithm used in this paper reached 95.7%.Then,by establishing the fit evaluation equation to calculate the f-value of different sizes,the correspondence between the sample body type and size was obtained,and the distribution pattern of each subdivision of body type in the size series was counted;and the regression equation was established to calculate the values of the control parts of each body type,and under the national standard 5-4 size series,the grade difference of different body types was obtained,and the body surface angle variables were divided into three categories to represent the different forms of each part respectively.Then we use radial basis neural network regression prediction algorithm to build a number recommendation model,and establish a radial basis neural network based clothing number recommendation system,explain the modules included in the system,and show the interface and operation process.Finally,for a larger number of body types with flat chests and average waist tucks,we design a series of garments that make use of the effect produced by the silhouette and pleats of the garment to make the overall form visually more beautiful.In summary,this study refines the human body type and clothing size on the basis of national standards,and uses the neural network method to construct a clothing size recommendation system,which provides reference for clothing tailoring and clothing size recommendation.