Automatic pavement detection based on video or image acquisition has become a common means of modern pavement management. However, the effect of pavement image processing is not quite well understood due to large amounts of noise. The low portion of crack pixels to the entire image makes the situation even worse. In this research, an efficient way of image preprocessing was developed for pavement crack images. This study included three steps, background correction, Gaussian smoothing and histogram transformation. Actual pavement crack images were preprocessed by the new method and then segmented by Otsu method. The results show that the preprocessing steps presented in this study dramatically dampened the impact of noise on image segmentation and retained most of the distress details at the same time. The Otsu method provided good segmented crack images if the images have been properly preprocessed. In the case of the crack images tested in this study, the preprocessing effect in the line cracking images was good enough for the following segmentation, and the effect in the alligator cracking images was even better.
展开▼
