With the linguistics as the research method, the work aims to construct a perceptual space of product modeling image and explore the characteristics and differences of perceptual expression of different attribute groups. Perceptual words are collected through literature review and association methods. The syllable method and slot method are used for filtering. Principles for selecting synonymous words and fittingness filtering are established. Based on the word frequency data, the influence of different attributes on people's perceptual expression is analyzed. Professional and non-professional backgrounds lead to significant differences in the selection of synonymous words and fittingness filtering. The professional participants tend to use professional, written, comprehensive, and precise words to describe product shape characteristics, while non-professional participants select significantly fewer words at the instinctive and reflective levels, more at the behavioral level, with relatively lower expression willingness and ability. Other attributes have little influence on the subjects' choices. It indicates that in the construction of perceptual space, the differences in perceptual expression needs of designers and consumers should be comprehensively considered to improve the quality of the perceptual space.