Abstract
Damage to road surfaces in the form of cracks and potholes increases over time and is compounded by poor maintenance systems. Potholes in particular can cause serious problems, including flat tires, damaged wheels, and car accidents. In a previous study, a pothole detection algorithm that used features of two-dimensional images to detect potholes was developed accurately. However, the algorithm yielded wrong detection in the case of similar objects, such as patches, stains, and shades. In particular, complicated shapes and random variations of similar objects led to misdetection. In this study, a pothole detection algorithm is proposed; it uses motion and the intensity of features to distinguish potholes accurately from similar objects. The motion feature is the source of primary information in the proposed algorithm and provides clear and noise-tolerant data for the extraction of potholes from the background region. The proposed algorithm consists of two steps of segmentation and decision and is much simpler than the authors’ previous method. Experimental results show that the proposed algorithm outperforms prevalent pothole detection algorithms as well as the algorithm from the previous study.
Get full access to this article
View all access options for this article.