Design Method Of Automobile Form Generation Based On Perceptual Image And Parametric Modeling

Electrification,networking and intelligence have become the development trend and trend of the whole automobile industry.In the contemporary era of serious technical homogenization,the automobile modeling design in line with users’ preferences is one of the key factors for automobile enterprises to win in the highly competitive market environment.The traditional automobile design and Research & Design cycle often takes several years,which is difficult to match the rapidly changing market demand,and usually depends on the designer’s personal inspiration,subjective experience and intuition.With the development of artificial intelligence technology,how to use new technology to quickly design a car that conforms to the user’s style image preference is a research direction worthy of exploration.Based on the generative design theory,combined with automobile style image,perceptual engineering theory,deep learning technology and parametric modeling technology,this paper studies an automobile form generative design method based on automobile style image,deep learning and parametric modeling,It provides a new design idea and method for designers to quickly explore the car form that conforms to the user’s style image preference and takes into account the family design characteristics of the brand in the conceptual design stage.This method mainly includes three parts: Kansei engineering,deep learning and parametric modeling.(1)In Kansei engineering part,this study extracts the top type side view curve with the highest relevance to modeling from complex products,establishes the mapping relationship between the image style dimension of perceptual engineering and the side view curve,carries out systematic cluster analysis on the data through the style image cognition experiment,and obtains the category labels of perceptual words of each sample.(2)In deep learning part,the deep learning model VAE-GAN is used to generate side view top profile image samples.Combined with the category labels obtained from the automobile style image cognition experiment,the automobile side view top profile deep learning data set is established,and the side view top profile image samples based on each style image vocabulary are generated through VAE-GAN.(3)In parametric modeling part,based on the parametric modeling platform,the automobile two-dimensional feature line is extracted and coded,the obtained automobile threedimensional feature curve is parametrically reconstructed,and the parameter correlation model of automobile three-dimensional feature curve driven by side view top profile is established.Based on the research results of the automobile style image perception experiment,the generation of side view top profile based on deep learning and the parameter correlation model of automobile three-dimensional characteristic curve,taking the side view top profile image generation results of each style image vocabulary as the input variables of the automobile threedimensional characteristic curve parameter correlation model,a large number of automobile forms with different forms and in line with the psychological cognition of users’ style image are generated,and the satisfaction of the generation results is investigated,The effectiveness of the vehicle form generation design method proposed in this paper is proved.