Product image conveys customers' emotional needs. It is the first time to use the method of multi-experiment to create the model of product image cognition model based on feature-matching, such as experts visit, eye track observing experiments of influencing the car's appearance features, the features extraction experiments influencing product image cognition, the quantitative relation experiments between key formal features and product image cognition. This research creatively takes the fuzzy relations as the most basic starting point, instead of membership functions. It adopts multi-matrix to describe the fuzzy relations qualitatively and quantificationally, and adopts random sets theory especially nested random sets to establish uncertain inference method. It has established the product image cognition model based on nested random sets. The method of product image feature lines creating technology based on possibility distribution and the method of the imprecise concept creating technology base on constraints are put forward. Furthermore, the computer-aided automobile sketch design system based on image cognition model is implemented. Through this system, customers can input the image words to get the related conceptual design sketches. Designers can make a sketch of simple formal features in the computers and the system can automatically bring forward the image corresponding to the sketch.