Automatic pavement crack detection based on image recognition

The traffic jam influences lots of losses socially and economically. One of main reason of traffic jam is caused by the maintenance of the pavement. It is better to detect and repair cracks on time to avoid big damage to pavement Therefore, lots of researches have been started to identify the crack with several methods. One method which is widely used is damage detection using image processing. Due to various complicated road conditions, such as oil pollution on the surface of the pavement and image noise, the classical algorithm cannot meet the detection requirements, the cracks cannot be extracted well. To solve the problems above, first, this paper introduces the mean filtering method to remove the image noise preliminarily and histogram equalization method to enhance the contrast of pavement image. Then, a pre-treatment before threshold segmentation was proposed to remove shadows and image noise. Then the crack is extracted by Niblack segmentation method and the image noise is filtered by removing the small area of the connected component. Finally, the crack type is classified using SVM (Support Vector Machine) and the location is obtained using the timeline to match GPS information and image.