Research On Product Image Design Method Based On Big Data Mining Of User's Perceptual Demands

Analyzing the user's kansei images and demands and incorporating them into the product image design can make the product more satisfy the user's kansei needs.For the current questionnaire-based user kansei image and appeal analysis methods,there are shortcomings in sample representativeness,reuse rate,and cognitive differences,and the kansei image and appeal analysis method based on text mining has the problems of insufficient utilization of text data and low support for design.In response to these problems,based on kansei cognitive theory,kansei engineering,and text mining technology,proposed a product image design method based on user's kansei appeals for big data mining.First,obtain product online evaluation text big data through web crawler,word2 vec is used for vectorized vocabulary representation,and the user's kansei image is divided into two categories: the overall kansei image of the product and the local kansei image of the product,respectively,through a graph-based keyword extraction algorithm and a keyword extraction method combining word vectors and syntactic relationships to extract kansei image and appeals,using the word vector model to represent it as a real number vector and performing principal component analysis for dimension reduction,combined with the importance parameter to calculate the corresponding kansei image parameters,to achieve the multi-level kansei image extraction method based on text mining.Subsequently,the product feature pictures are scaled and drawn using parametric curves.Product features are divided into two types: global features and local features.Discrete representations of the morphological proportional parameter based on the position of the hard points,and the numbering of the local design features are used to extract the product features.Then,through the maximum information coefficient,the dimensional selection of the kansei image parameters is carried out.Based on the random forest classification algorithm,a kansei engineering expert system is constructed,and the multi-layer kansei image parametric extraction of the user's kansei demand text is input into the kansei engineering expert system to output design feature parameters,so as to draw the generation of new design schemes and realize the product image design method driven by kansei image big data.Finally,based on the case of the side styling design of the sedan,the method was practiced and verified.The results show that the proposed multi-level kansei image extraction method based on text mining can quickly and targetedly analyze the user's kansei image and realize the parameterized representation.The proposed product image design method driven by kansei image big data can satisfy consumers' kansei appeal to a certain extent.The method proposed in this paper can quickly obtain users' kansei images and demands from a large number of online evaluation texts,which avoids the shortcomings of the questionnaire method.While obtaining the perceptual image of the whole and local features of the product,it can parameterize the expression of kansei image without the limitation of the number of words.At the same time,it supports for adding new image vocabulary,which improves the utilization of text data.Users' kansei knowledge in the text can be applied to the design of new products without case-based reasoning,which greatly reduces the manual operation steps and time required,improves the supporting role of text mining on the design and design efficiency.