Improving Classification Accuracy in Paint Surface Inspection Using Image Processing and Deep Learning

In recent years, research has been conducted on defect detection and classification using image processing and deep learning as alternatives to visual inspection. This paper proposes a method to detect defects by applying embossing and histogram equalization to images of painted surfaces and classify defects using ResNet50. The proposed method is capable of simultaneously classifying multiple defects in an image. Furthermore, we show that the proposed method can improve the accuracy of defect extraction compared to the conventional method.