Research On Design Of Cheongsam Structure Based On Neural Network

Cheongsam,as a traditional Chinese dress,is the first choice for ladies to attend grand occasions,but the fit of cheongsam when buying is often unsatisfactory.In the market,most cheongsam fabrics are made of silk or cotton,and the fabric itself has limited elasticity.The non-stretched slim cheongsam is difficult to meet the needs of various body types,so most people will choose personalized customization.However,personalized customization requires one person and one board,which takes a lot of time and cost,which is not conducive to the long-term development of the enterprise.In order to solve this problem,this paper proposes a GA-BP neural network mapping model that uses the anthropometric size to the detailed size of the fitted cheongsam model.Through the secondary development of Auto CAD,an automatic output system for the cheongsam model is established to export cheongsam structures of different sizes Model to improve the speed of the enterprise’s production of fit cheongsam.Firstly,this paper introduces the current situation of garment customization companies and analyzes the difficulties they are facing.A large number of relevant literature on neural networks,related research on cheongsam models was analyzed and summarized.The non-contact three-dimensional measuring instrument was used to collect human body data,and the 179 groups of human body data were finally retained after analysis by SPSS software.According to the needs of the experiment,the detailed size of the cheongsam and the human body data were collected as training data samples,and the neural network was trained later.Secondly,the derivation model of GA-BP neural network is constructed.Set 9 human body size parameters as input values,22 fit cheongsam template detail sizes as output values,13 hidden layer nodes,hidden layer transfer function as tansig function,output layer transfer function as purelin function,network training function is the trainlm function,which uses a genetic optimization algorithm to select and optimize the randomly obtained thresholds and weights,fully combine the 179 sets of data obtained in the previous period,select the previous155 sets of data as the research sample,and select the following 24 sets of data as the test Data,and use Matlab to build GA-BP network.After 83 iterations,it shows that the fitness is 0.9884,the error between the sample predicted value and the actual expected value is small,and the expected effect is achieved.Then,through the second development of Auto CAD through the LISP development environment,a fitting cheongsam model automatic generation system is constructed.The system is divided into two parts,namely the size input module and the model generation module.The size input module creates a detailed size input interface,and automatically generates acheongsam template based on the relevant parameters of the front piece,back piece,collar piece and sleeve piece of the cheongsam structure template.Finally,the automatic generation of cheongsam models was evaluated from pattern-making theory,three-dimensional virtual fitting and clothing fitting.The comprehensive analysis results show that the automatically generated cheongsam fitting models meets the requirements of human body for aesthetics and comfort,and it has high practicability,and can be applied to actual production and life.