Abstract: In order to deal with the different surface textures and various types of automotive brake pads, an image detection method for surface defects of brake pads based on machine vision was proposed, which combined the gray co-occurrence matrix and the density clustering. Firstly, the friction surface of the brake pad was extracted. Then the construction factor selection of the gray co-occurrence matrix was determined, and the linear correlation was used to select the characteristic parameters. Each friction surface was divided into several small windows, and the eigenvalue of each small window was calculated to construct the feature data sets. The cluster analysis was carried out by the adaptive density clustering algorithm, so as to further determine whether there existed a defect area or not. Based on the statistical analysis of the test results of 58 brake pad samples of different models, the experiment shows that the method has a high detection accuracy with a false detection rate of 8% and a missed detection rate of 6%. Therefore, the method can better detect whether there are defects in the brake pad samples and has a wide range of applicability.