A novel method of stone surface texture image recognition

Abstract

With the development of stone processing and sales, effective stone surface texture image recognition methods are needed. We proposed a new stone surface texture image recognition method based on texture and colour. We combine the following visual features: Gabor features which can well simulate the single cell sensing profile of mammalian visual neurons, The Grey-level Co-occurrence Matrices(GLCM) which describe image gray distribution characteristics and spatial location information, and HSV colour features which are consistent with human visual characteristics. In addition, for the sub-image of the stone surface texture image can contain its original image texture structure, this paper adopts the block training idea, subdividing original image into non-overlapping sub-images to multiply the number of training samples for SVM classifier. Extensive experimental results show that the proposed method has a reference value for the study of stone texture image recognition.