Implementation of K-Nearest Neighbours for Automotive Body Surface Defect Detection

A body repair shop company that handles surface damage to vehicles is very necessary because the exterior appearance of the vehicle plays an important role in creating a good and attractive impression to others. Therefore, body repair companies must be able to adapt and keep up with the times. By combining advanced technology solutions, such as image processing, companies can streamline the identification phase, reduce human error, and optimize the allocation of repair resources. The problem discussed in this study is how to classify the surface of the vehicle body so that the classification model built can increase the accuracy in classifying the surface of the vehicle body. The steps taken in this study are to collect images of the surface of the vehicle body, and then the image data goes through a preprocessing process using median filtering to remove noise and segmentation techniques to improve image results. After preprocessing, the next step is to extract features based on texture using the Gray Level Co-occurrence Matrix (GLCM) and Statistical methods. Next, the images will be classified using the K-Nearest Neighbors (KNN) method, and the accuracy obtained is 56.80% for k = 3. After that, the classification model will be evaluated using the Area Under the Curve (AUC), and the AUC value obtained is 68.75%. With this approach, body repair workshop companies can improve the efficiency and accuracy in classifying vehicle body surfaces by utilizing image processing technology and techniques, allowing for better resource allocation and more reliable repair results