Abstract

Cracks are the most significant type of pavement disease, and the precise segmentation of cracks serves as an important decision-making basis for national preventive road maintenance management. In response to the problem of crack segmentation accuracy of existing pavement models under complex backgrounds, a crack recognition algorithm for remote sensing images based on the Fuzzy Automatic Threshold C-Means Clustering Algorithm (FATCM) was designed by incorporating local spatial and gray-level information constraints. The FATCM method can strengthen the inherent effectiveness of the traditional fuzzy C-means (FCM) algorithm, achieve uniform segmentation through fuzzy membership calculation and iterative process, and effectively eliminate edge ambiguity. The core innovation of FATCM resides in the introduction of the fuzzy local similarity measure, which is predicated upon the pixel spatial attraction model. This novel measure is astutely applied to automatically strike a refined equilibrium. Specifically, it ensures a high degree of insensitivity to noise, a factor of paramount importance in safeguarding the integrity of image data. Simultaneously, it minimizes the manifestation of edge-blurring artifacts, thereby proficiently retaining the minute and crucial details of the image. Multiple types of images in the Crack500 dataset were used in the experiments to evaluate the performance of FATCM. The experimental results show that this method has good detection results and can effectively extract weakly contrasted cracks and small cracks.OPEN ACCESS Received: 24/07/2024 Accepted: 16/12/2024 Published: 20/04/2025