A Proposed Intelligent Map Fusion for Pavement Crack Detection Using Structure from Motion

Due to its small dimension and low-intensity variation to the background, crack is considered the most challenging distress on the pavement surface. Despite multiple methods deployed to detect and classify cracks, Smart City still intends for an intelligent and cost-effective solutions. This paper proposes a novel approach based on an intelligent Map Fusion of extracted features from 2D images and 3D models. The proposed method transforms pavement images taken by a commercial camera into a 3D model using the Structure from Motion as a computer vision technique. Then, the grayscale level and roughness are computed for each 3D point and filtered to produce the crack 3D Point candidates at the decision-making level. Finally, the Graham Scan algorithm was used to define the contour of the crack. The presented method has been assessed and validated for a case study. The findings indicate that the model can detect and accurately contour the crack.