A Color Segmentation and Feature Matching Algorithm for Car Brake Pad Image Classification

Abstract

During the manufacturing of car brake pads, classification of brake pads with different Appearance is an important process, which at present is done mostly through the human eye detection. In this paper, an intelligent car brake pad image classification algorithm based on machine vision technology is proposed. Firstly by using the high-resolution industrial camera with coaxial light source, the image of brake pad on the conveyor is acquired. Then a HSV color space conversion is conducted on the original image. By thresholding the H or S channel of the image and morphological processing, the brake pad foreground is accurately segmented. Following the segmentation, the Hu moments of the brake pad region is extracted as a shape descriptor of the pad. At last the 7-dimentional Hu feature is compared to the templates of different kinds of brake pads to find the best match. Experiments show that the proposed algorithm can successfully segment the brake pads from the dark indistinct background and the classification accuracy reaches 81.7%.