Improved Road Crack Detection Utilizing Pixel Categorization with Linear Relationship Based Augmentation in Robust Fuzzy-C Means Clustering

Abstract

Roads with many cracks are dangerous, hard to inspect manually and required extensive repairs if left unaddressed. Automating crack detection can save time and money, but it's difficult due to poor image quality. To address this, we present a powerful and novel Fuzzy C-Means clustering method for automating fracture identification. This approach utilizes a 3×3 window that encompasses the whole picture and then categorized the pixels into edge or non-edge pixels using a second order difference equation prior to segmentation. Moreover, it allows for edge pixel augmentation within every window, which effectively highlights the details of fractures. This enhancement employs an augmented scaling factor derived from pixel contribution ratio alongside Michelson contrast to improve the edge and crack detection accuracy. Furthermore, the intensity difference is incorporated to addressing the ambiguity that arises in cluster assignments when Euclidean distances are identical during segmentation, leading to more precise and reliable fracture identification. Additionally, the proposed novel algorithm demonstrates effective crack detection on unfamiliar photographs across various scenarios, without the reliance on a training dataset. The empirical findings indicate that the proposed Fuzzy C-Means Clustering algorithm (called as CLAFCMC) achieves superior performance in term of Partition Entropy, Davies-Bouldin Index, and Partition Index values compared to the existed methods such as K-Means Clustering, Fuzzy C-Means Clustering, and Manhattan distance-based Fuzzy C-Means Clustering for road crack detection. Furthermore, the algorithm optimizes computational efficiency, significantly reducing execution time. These results validate the algorithm's reliability and effectiveness, positioning it as a highly promising solution for automated road crack detection systems.