A segment algorithm for crack dection

Abstract

Proposed a robust and high-efficiency model of segmentation for pavement images which based on wavelets transform. As the crack pixels in pavement images are darker than their surroundings and continuous, images associated with the same road section with the consistence of pavement crack. The designed model contains two stages mainly: Image segmentation is implemented based on wavelet transform method. The wavelet decomposition coefficients are further transformed to remove noises and to enhance the edges of cracks. Finally, simple threshold technology is used to generate binary image and discriminate cracks against background interference. Experiments results demonstrated that the proposed method can detect the cracked images from massive pavement images correctly and effectively.