Abstract: Wood appearance affects consumers’ preference of products, and determines the price of products and sales to a certain extent. Colors and texture on the wood board are random and diverse, so using statistical methods to classify them has limitations. According to the characteristics of board surface, we propose a board classification method based on user’s visual characteristics of. This method converts user’s visual preferences into a quantitative analysis of the color and texture, then optimizes samples through mining features demonstrated by the sample data and status information, and finally designs a compression perception classifier for information integration and wood classification. For colors, Tamura texture features and the basic statistical features were used as feature vector, which include six parameters corresponding to six kinds of psychology properties. These features can fuse user’s visual psychology with the basic statistics, and give a more accurate and complete expression of visual features on the board surface. What’s more, the genetic algorithm was implemented to optimize training sample by its nonlinear mapping. For classifier design, compressed sensing was employed as classifier, an over-complete dictionary was built by L*a*b* color feature, Tamura feature and the basic statistical characteristics, and the classification result was obtained by solving the minimum of l1 norm to find out the characteristic data of the test sample and the class vector in the over complete dictionary. Experiments show that the accuracy is 90.56% and the sorting method is practical.