ASPHALT PAVEMENT CRACK DETECTION USING IMAGE PROCESSING AND NAÏVE BAYES BASED MACHINE LEARNING APPROACH

Abstract

For appropriate road maintenance and management of the road, detection and evaluation of the asphalt distress is very important. Standard test method for asphalt pavement in Japan have defined that crack ratio, amount of rutting, and flatness are main damage indices for the evaluation. Though the amount of rutting and the flatness can be obtained automatically from the pavement condition survey vehicles, the crack ratio is not automatically derived. Since this process requires to draw and count cracks manually from the photo of the road surface taken by the vehicle, enormous effort and time are required. In addition, the accuracy of the evaluation is doubt because the manual method does not evaluate the important information including the crack width. To solve these problems, this paper have developed automated asphalt pavement crack detection method using image processing technique and naive Bayes based machine learning approach. The developed method is tested using the photos of various types of asphalt pavements, and it is found that the method can detect cracks with very high accuracy.