Surface Defect Detection on Aluminum Profiles After Roll Bending Based on an Improved Fuzzy Clustering Algorithm

This study is dedicated to defect detection on the surface of aluminum profiles after roll bending processing, which is crucial for ensuring the quality, safety, and reliability of aluminum profile products. In view of the limitations of traditional manual visual inspection, a detection method based on the improved Fast Robust Fuzzy C-Means (FRFCM) theory is proposed. The method primarily consists of three steps: Initially, image preprocessing is conducted, which involves grayscaling and filtering operations to optimize image quality and suppress noise interference. Subsequently, the preprocessed image is segmented using the FRFCM method specifically on the surface defect images of aluminum profiles, aiming to preliminarily distinguish between defective and non-defective areas. Finally, edge detection is performed, utilizing the Canny edge detection operator to further process the preliminary segmentation results, thereby accurately extracting the edge contours of the defect areas on the aluminum profile surface. During the image segmentation stage, a systematic comparison was conducted among four segmentation methods: Otsu's thresholding method, K-means clustering algorithm, Fuzzy C-Means clustering algorithm, and FRFCM algorithm. Experimental results demonstrate that the defect detection method based on FRFCM exhibits optimal performance in terms of the Jaccard Similarity Coefficient index, ensuring high accuracy and reliability, and possessing significant advantages in terms of reproducibility and efficiency. The proposed segmentation and detection method based on FRFCM provides robust support for surface defect detection in aluminum profiles after roll bending and holds significant importance for quality control in aluminum profile production.