Computer vision for road imaging and pothole detection: a state-of-the-art review of systems and algorithms

Abstract

Computer vision algorithms have been utilized for 3-D road imaging and pothole detection for over two decades. Nonetheless, there is a lack of systematic survey articles on state-of-the-art (SoTA) computer vision techniques, especially deep learning models, developed to tackle these problems. This article first introduces the sensing systems employed for 2-D and 3-D road data acquisition, including camera(s), laser scanners and Microsoft Kinect. It then comprehensively reviews the SoTA computer vision algorithms, including (1) classical 2-D image processing, (2) 3-D point cloud modelling and segmentation and (3) machine/deep learning, developed for road pothole detection. The article also discusses the existing challenges and future development trends of computer vision-based road pothole detection approaches: classical 2-D image processing-based and 3-D point cloud modelling and segmentation-based approaches have already become history; and convolutional neural networks (CNNs) have demonstrated compelling road pothole detection results and are promising to break the bottleneck with future advances in self/un-supervised learning for multi-modal semantic segmentation. We believe that this survey can serve as practical guidance for developing the next-generation road condition assessment systems.

A modified pothole detection approach to capture their width

“…In this section, we propose an improvement of the long-established pothole detection method given by Koch and Brilakis [2]. The chosen approach is one of the first works addressing the computer vision problem of detecting potholes [1]. They proposed a two-stage 2D image processing model that can be applied directly to a vehicle's rear camera.…”

“…Efforts to mitigate pothole problems have seen a rapid increase in number, with a wide variety of computer vision-based techniques dominating the field of automatic detection. They are classified into three main types, namely classical 2D image processing, 3D point cloud modeling and segmentation, and machine learning/deep learning [1]. One of the earliest works was developed by Koch and Brilakis [2], in which they used histogram shape-based thresholding to classify defect and non-defect regions, and then compared the interior texture to decide whether the region of interest truly represents an actual pothole.…”

Computer Vision Applications in Intelligent Transportation Systems: A Survey

“…State-of-the-art systems and algorithms for road imaging and pothole detection were investigated by Ma et al [ 345 ]. In this research, (i) classical 2D image processing, (ii) 3D point cloud modeling and segmentation, and (3) ML/DL methods for pothole detection were analyzed in detail.…”

Deep Learning-based Pothole Detection for Intelligent Transportation: A YOLOv5 Approach

“…These cities leverage cutting-edge infrastructure, digital connectivity, and intelligent systems to enhance the quality of life for their residents [2,3]. The deployment of smart city facilities enables efficient management of resources, optimized transportation systems, and improved public services [4,5]. However, amidst these advancements, the issue of deteriorating road conditions, particularly potholes, remains a persistent challenge.…”

“…In recent years, several technologies and methodologies have been proposed for pothole detection. Vision-based methods have gained a lot of attention because of their nondestructive nature and real-time applicability [4,8,9]. Visionbased techniques utilize cameras and image processing algorithms to analyze road surface images and identify potential potholes [10,11].…”