The U.S. DOT John A. Volpe National Transportation Systems Center (Volpe), under the direction of the Federal Railroad Administration (FRA) Office of Research, Development, and Technology, studied the use of drones to produce accurate 3-dimensional models of high-profile highway-rail grade crossings. Volpe found that photogrammetry with ground control points can produce models with similar accuracy to those products using LiDAR at a much lower cost. Volpe determined the process needed to achieve a simple measurement of ground clearance for these crossings, enlisted the advice of DroneUp, and made recommendations to the FRA for pursuing this capability.
While there are guidelines for constructing new highway-rail grade crossings aimed at preventing vehicles from becoming stuck on the tracks, there is currently no formula or threshold for determining if an existing crossing presents a risk for low ground clearance vehicles, or if it should be posted as such. It is left to states and municipalities to determine the risk through whatever means they choose. This study chose drones.
The objective sought to determine the viability of using drones to capture imagery that can yield accurate point cloud models of humped railroad crossings. It aimed to determine the pros and cons of photogrammetry (using photographs to produce point cloud data) versus those of LiDAR.
To better understand the pros and cons of both photogrammetry and LiDAR, Volpe issued a contract to DroneUp to produce point clouds of a humped crossing using both technologies. Through this approach, Volpe did not need to purchase its own aerial LiDAR system, plus it could gain the insights of experts in both technologies regarding the capture and back-end processing challenges.
This study investigated the viability of using drone photogrammetry to capture accurate point cloud data of high-profile railroad crossings that present a hang-up risk to low ground clearance vehicles. It also aimed to explore the strengths and shortcomings of aerial LiDAR as compared to photogrammetry. Finally, this study examined how best to use this information to improve safety at high-profile railroad crossings.
What is photogrammetry, and what are the software needs? Photogrammetry is a process for creating 3-D models from a series of interrelated photographs. While this is not a new process, the recent emergence of inexpensive, high-quality drones, coupled with the development of modern modeling software tools, has resulted in a new capability for creating highly accurate models with relatively little time and effort.
Volpe had an existing perpetual software license to perform trespasser detection. This software takes drone imagery and, with the help of the GPS data captured with the images, creates 3-D models from detailed point clouds.
The software can use any imagery to create point clouds, including videos and photos taken from cell phones. However, best results are achieved when individual images are captured with a drone because these also capture the GPS data for each picture, which helps the software associate the images with one another. Creating point cloud models using LiDAR scanners has been an accepted practice among mapping and surveying professionals for several years. More recently, small, lightweight LiDAR sensors have been developed for use on drones.
Both photogrammetry and LiDAR data captured from drones are capable of producing highly accurate three-dimensional models of high-profile grade crossings. However, LiDAR models require equipment that is much more expensive, and processing the data is far more labor intensive than photogrammetry as found in the USDOT FRA detailed report.