Road potholes have a well-known impact on driving quality and safety. Therefore, timely mitigation of potholes is critical for the safety of road users. However, efficient and timely maintenance relies on the presence of an effective process for pothole detection. Currently, transportation agencies primarily rely on manual inspection and road user reporting. These methods are subjective, prone to inaccuracy, and some are also laborious and time-consuming. An ideal pothole detection system would be accurate, objective, automated, and relatively inexpensive. In this context, accuracy encompasses three distinct performance areas: detection, localization, and size estimation. This study explores the potential of utilizing a mobile light detection and ranging (LiDAR) for accurate detection and size estimation, along with a global navigation satellite system (GNSS) receiver for localization, to develop an effective pothole surveillance system. To achieve this objective, the study proposes a four-step framework. Firstly, the LiDAR data are processed to generate ring-wise cross-sectional images. Secondly, a deep learning object detection network is trained to predict the presence and size of potholes. Thirdly, the ring-wise inferences are aggregated to produce a final decision. Lastly, the aggregated inferences are synchronized with GNSS locations to generate inspection maps. The system’s performance was validated using multiple road strips, never seen by the model, containing potholes of different sizes and shapes. The results demonstrated the effectiveness and accuracy of the proposed system. Overall, this research contributes to the research on LiDAR-based pothole inspection by proposing a novel four-step framework and incorporating it into an end-to-end pothole detection system, which can greatly improve the efficiency of pothole maintenance and enhance the safety of road users.
Roadside incidents are a leading cause of driver fatalities in the United States, with a significant number involving collisions with barriers, such as guardrails. Guardrails are essential safety barriers designed to maintain vehicle trajectories and shield against roadside hazards. The functionality of guardrails heavily relies on their structural integrity, and damaged guardrails can pose serious dangers to road users. Traditional inspection methods are labor-intensive, time-consuming, and prone to human error, lacking periodic monitoring crucial for timely maintenance. Although advancements in computer vision have enabled automated infrastructure inspections, research dedicated specifically to the inspection of guardrails remains scarce. Existing automated solutions do not fully address the challenges of accurately identifying and assessing guardrail damage under varying lighting and weather conditions and the computational demands of real-time processing. This study addresses these challenges by introducing a novel framework utilizing advanced computer vision techniques, such as YOLOv8 models and the Deep OC–SORT tracker, integrated with camera and GPS systems mounted on a vehicle. This system automates the detection, localization, and severity assessment of guardrail damage, enhancing inspection accuracy and efficiency, enabling faster maintenance responses, and ultimately contributing to safer road conditions.
21 May 2024 in Transportation Research Record: Journal of the Transportation Research Board
Given the potential hazards and risks that potholes pose to road users, this study introduces an image-based system that utilizes a combination of a camera and GPS for real-time detection, georeferencing, and area estimation of potholes. The captured system data is processed in real-time using YOLOv8, a deep learning model proficient in object detection and segmentation. To enhance the precision and reduce the occurrence of false detections, the system is specifically trained to detect potholes, manholes, and patches. Additionally, the camera is calibrated to accurately estimate the area of identified potholes. The proposed system achieved a mean average precision of 91% in detecting potholes, 98% in detecting manholes, and 90% for detecting patches. A salient feature of this system is its capability to localize potholes with reference to pavement line lane markings. This ability could facilitate proactive lane closure planning by maintenance crews, further enhancing road safety measures. The study findings suggest that the system holds significant potential for practical implementation. Its deployment could assist transportation agencies in the prioritization of road repairs, resource allocation, and advance planning for lane closures, ultimately enhancing the efficiency of their maintenance workflows.
Given the crucial importance of pavement marking retroreflectivity in ensuring visibility for road safety, this research investigates the correlation between pavement marking reflectivity and LiDAR data. Empirical data were collected from eight road sections using both a handheld retroreflectometer and a mobile LiDAR. The approach proposed focuses on extracting important features from pavement marking regions of the LiDAR point cloud. A comprehensive feature extraction and feature selection process was employed. In addition, a well-rounded selection of learning algorithms was evaluated. A rigorous hold-out evaluation was incorporated, ensuring that the reported performance metrics were robustly generalizable. The best performing model was able to achieve an R2 of 0.824 on unseen data. The findings of this study illuminate the potential for leveraging relatively inexpensive mobile LiDAR sensors in combination with machine learning techniques in conducting efficient pavement marking assessments, not only to detect completely degraded markings, but to accurately estimate retroreflective properties.
24 June 2022 in Transportation Research Record: Journal of the Transportation Research Board
Road edge drop-off has well known impacts on the safety of the traveling public and roadway service life. Continuous inspection of roadsides is therefore needed to monitor their condition. Autonomous vehicles (AVs) are increasing in number and have the potential to be used for road and roadside condition data collection. This study explored the use of AVs to assess road edge drop-off. To this end, the ability and accuracy of a research grade autonomous vehicle platform built on a passenger car was used to determine road edge drop-off. Data were collected for the roadside along a state highway in Ohio using the different sensors of the vehicle platform, including a Light Detection and Ranging (lidar) sensor. A 160-m (525-ft) long section of the highway was selected and surveyed using a high accuracy stationary terrestrial laser scanner to obtain the topographic map of the highway and its sides. The lidar data were analyzed using fully automated deep learning methods to determine the edge drop-off severity along the selected section. The analyzed lidar data were compared with those obtained using the high accuracy stationary terrestrial laser scanner. The results of this study showed that the autonomous vehicle platform can be used successfully to assess the road edge drop-off with excellent accuracy, particularly when using end-to-end deep learning methods for analyzing the collected data.
24 April 2025 in Transportation Research Record: Journal of the Transportation Research Board
Traffic-sign condition assessment is a critically important responsibility for transportation agencies. This paper presents an innovative framework to address this crucial task by leveraging mobile sensors, a machine vision camera, and a Global Positioning System (GPS) for real-time traffic-sign condition assessment. An artificial intelligence (AI) application, integrated into a robotic operating system framework, processes image and GPS data streams. The system deploys the YOLOv8 computer vision algorithm for traffic-sign detection and segmentation, supported by ByteTrack and a custom correction algorithm for object tracking. This sophisticated system segments traffic signs from other objects, enabling shape- and color-based damage assessment. Distance and speed from a GPS is also employed to calculate the stopping sight distance to identify severe obstructions. The study successfully demonstrated the efficacy of this approach by integrating these components into a seamless framework capable of real-time segmentation, classification, and geo-referencing of damaged traffic signs at regular traffic speeds. The XGboost machine learning algorithm, chosen for its accuracy and real-time prediction time, was used for second-stage damage classification. This work validates the value and effectiveness of AI and mobile sensor technology in traffic-sign condition assessment, making it a scalable, cost-effective solution for transportation agencies. Future work will seek to broaden the model’s capabilities to detect a wider range of traffic-sign types and conditions and to validate the effectiveness of the framework in real-world environments.