Analysis of Correlation Between Image Texture and Friction Coefficient of Materials

It is unknown that whether friction coefficients of materials can be predicted by their images. In this paper, we explore the correlation between the image gray-level and the friction coefficient of materials. We introduce a systematic approach to find the correlation model. First, four key features were extracted from Gray-Level Co-occurrence Matrix (GLCM) using Hue Saturation Intensity (HSI) color space. Second, BP neural network was utilized to establish the correlation model between the image gray-level and the friction coefficient. The proposed approach was validated using a dataset with 100 samples. The results show that the average regression error of the model is 16.7% for the 100 samples, and 2.8% for the subset of 30 fabric samples among the totals. Within those fabric samples, the prediction error for new samples is 20.1%. The experimental results indicate a possibility of inferring the friction coefficient from the image of the material. This study might provide a way of automatically constructing a haptic database through the large amount of images on the internet.