Crack detection in shadowed images on gray level deviations in a moving window and distance deviations between connected components

Abstract In the detection of pavement cracks in an image, shadows often affect the detection result seriously. To extract the cracks accurately and effectively from shadowed pavement images, a method including a number of algorithms and functions is studied based on the gray level standard deviation in a local window and the distance standard deviation in a connected region, which is different to the traditional algorithms/methods based on image processing. The proposed framework begins with selecting a moderate sized window automatically according to the resolution of the treated image. Then the pavement image can be segmented by a threshold determined by the mean value of gray level standard deviation in the window. Subsequently, the crack segments can be extracted using the distance standard deviation of the connected components. Finally, the segments can be connected according to the gap lengths and segment direction information. We tested about 300 pavement crack images in which the shadows are caused by trees, buildings, grass, telegraph poles, street lamps and so on, and we compared the new method to more than ten different traditional algorithms/methods such as different edge detectors, Thresholding, Minimum Spanning Tree, Clustering analysis and FCM algorithms. The testing results show that the new method for the pavement crack detection in different shadowed images is satisfactory, the detection accuracy can be up to 96%, and the algorithm comparison proves that the proposed algorithm is much better than that by the widely used traditional algorithms.