Analysis for existed pavement cracks extraction algorithm, pointing out the problems and deficiencies, a phase-grouping pavement cracks extraction method based on genetic algorithm was proposed. This method first used a genetic algorithm to determine the optimal threshold, which achieved the threshold of image segmentation; then adopted phase-grouping method to extract pavement cracks information. Experimental results show that the continuous cracks can be stable and effectively extracted, and the computational complexity is linear, it is conducive to high-speed real-time processing.