Fast and robust pavement crack distress segmentation utilizing steerable filtering and local order energy

Abstract

Pavement crack plays an important role in estimating pavement conditions and implementing pavement maintenance management. There are three shortcomings summarized. (1) Low contrast between crack pixels with surrounding pixels; (2) intensity values along crack discontinuous; (3) complexity structure instead of unitary and unbranched cracks. The proposed algorithm, including three essential steps to address the above disadvantages. A fast nonlocal means denoising method (FNLM) is first proposed to eliminate Gaussian isolated noise while retains target pixels. Second, the basic filter named steerable filter (SF), which calculates the maximal response with different orientations at a certain position, is made up of the second partial derivatives of a two-dimensional Gaussian function. Then the crack saliency map is obviously generated. Last, the local order energy (LOF) is applied to extract crack features from the saliency map to generation binary image, and the mathematical morphology operation method is utilized to locate the location and orientation of crack. Different types of pavement images are applied to estimate the performance of the proposed algorithm, and the experimental results represent the proposed method can achieve fast and robust performance.