Automated assessment of vertical clearance on highways scanned using mobile LiDAR technology

Assessing vertical clearance at bridges is a preliminary step in most routine bridge inspections. This information is critical when assessing the structural integrity of bridges. Furthermore, clearance information at bridges and other overhead assets on a highway network is also extremely important when routing oversized vehicles on a highway network. Efficient clearance assessment makes critical information readily available to asset owners. As a result, asset owners and transportation agencies are able to address concerns in a timely manner, which would help them avoid prohibitive maintenance costs sustained in case of collisions. Unfortunately, manual clearance assessment using conventional surveying tools is unsafe, time consuming, labour intensive. To overcome these challenges, this paper proposes a novel algorithm whereby mobile LiDAR data could be used to efficiently assess clearance at overhead objects on highways. The proposed algorithm first detects and classifies all overhead objects on a highway segment. The clearance is then assessed at each of those objects and minimum clearance is identified. Detection involves voxel-based segmentation of the point cloud followed by a nearest-neighbour search to locate overhead structures. After detecting the structures and identifying their locations, points representing the same object are clustered and classify into bridges and non-bridges. Furthermore, an estimate of the clearance at each object is also obtained. For objects of long span such as bridges, detailed clearance assessment is performed. The developed algorithm was tested on three highway segments in Alberta, Canada including a 242 km highway corridor. Testing revealed that the method was successful in detecting and classifying all overhead structures at an accuracy level of 97.8% and 96.2%, respectively. The algorithm was also successful in accurately measuring the clearance at those structures with the differences in measurement between ground truth data and the extracted results ranging between 0.03 and 0.47%.