Automatic Pavement Crack Detection Using Texture and Shape Descriptors

AbstractPavement distress detection and analysis is the most important part of automated pavement inspection systems. Due to the circumstances such as complex texture, uneven illumination, and nonuniform background, pavement distress detection is not a simply edge detection process. Over the past 30 years, lots of methods were proposed to detect pavement distresses, especially cracks. In this letter, a novel automatic pavement crack detection approach based on texture analysis and shape descriptors is proposed. Pavement surface is seen as a texture surface, and distresses are defined as inhomogeneities occurring in the texture surface. Six texture features and two translation-invariant shape descriptors were used here as discriminate features against irregular texture and uneven illumination. By using a SVM classifier, all sub-images are classified as crack or non-crack. Final results were obtained after post-processing, which includes segmentation, fake-crack eliminating, and crack-measuring methods. Com...