Abstract: This paper proposed a pyramid-based cracks statistical method for pavement images on the basis of the consistence of pavement texture structure.After the enhancement process,it executed image segmentation based on multi-scale space algorithm,then the weighted pyramid-based cracks statistical was implemented on each segmented image,in the end picked up the cracked pavement images and achieved the objective of improving detection efficiency.Several experiments were executed aim at road sections with different pavement texture structure with this model,and the results demonstrated that it can pick up the cracked ones from massive pavement images correctly and effectively.