College

College of Engineering & Technology (Batten)

Department

Civil & Environmental Engineering

Graduate Level

Doctoral

Presentation Type

Poster Presentation

Abstract

In this study, we fuse data from a 3D LIDAR mounted on a vehicle and images from an external traffic surveillance camera to create a 3D representation of a segment of a roadway that experiences frequent flooding. The point cloud data from the LiDAR in this study is collected from a road segment of W 49th Street in Norfolk, near the ODU campus. The traffic surveillance camera is mounted on a public parking building in the same area. The LiDAR collects continuous point cloud frames as the vehicle traverses the section. Multiple LIDAR frames related to the respective segment of the road, monitored by the external camera, are first merged into a unit point cloud using the ICP registration method, representing a local high-resolution digital elevation model (DEM) of the road segment. Then, the resulting DEM is projected onto the images of the flooded road captured by the surveillance camera. To this end, a camera calibration technique is employed to estimate the transformation parameters. The camera calibration method relies on a dataset comprising points and their corresponding pixels in the target image. A virtual grid of points and corresponding pixels was generated to run a camera calibration function. The mentioned dataset was generated with the aid of projecting point clouds on LiDAR’s internal camera, enabling us to identify objects and curbsides. The perspective geometry principles were also employed to create the DEM. The projection results show the successful performance of the employed technique for camera calibration. The depth estimation was carried out using the projected DEM model on a flood image recorded by the external camera.

Keywords

Urban Flooding, Depth Detection, Mobile LiDAR, Camera Calibration

Share

COinS
 

Estimating Depth Of Roadway Flooding Using Data From LiDAR And Surveillance Cameras

In this study, we fuse data from a 3D LIDAR mounted on a vehicle and images from an external traffic surveillance camera to create a 3D representation of a segment of a roadway that experiences frequent flooding. The point cloud data from the LiDAR in this study is collected from a road segment of W 49th Street in Norfolk, near the ODU campus. The traffic surveillance camera is mounted on a public parking building in the same area. The LiDAR collects continuous point cloud frames as the vehicle traverses the section. Multiple LIDAR frames related to the respective segment of the road, monitored by the external camera, are first merged into a unit point cloud using the ICP registration method, representing a local high-resolution digital elevation model (DEM) of the road segment. Then, the resulting DEM is projected onto the images of the flooded road captured by the surveillance camera. To this end, a camera calibration technique is employed to estimate the transformation parameters. The camera calibration method relies on a dataset comprising points and their corresponding pixels in the target image. A virtual grid of points and corresponding pixels was generated to run a camera calibration function. The mentioned dataset was generated with the aid of projecting point clouds on LiDAR’s internal camera, enabling us to identify objects and curbsides. The perspective geometry principles were also employed to create the DEM. The projection results show the successful performance of the employed technique for camera calibration. The depth estimation was carried out using the projected DEM model on a flood image recorded by the external camera.