EVALUATING THE ACCURACY OF LIDAR POINT CLOUD REGISTRATION VERSUS GROUND CONTROL POINT RECTIFICATION FOR COASTAL MAPPING

Quantifying coastal morphology on the time scales of minutes to hours before, during, and after extreme events (e.g., tropical cyclones and nor’easters) is necessary to understand how severe weather events impact our coastline. Such a measurement system must be (i) able to operate in extreme weather (e.g., strong winds and heavy rainfall), (ii) operable during both day and night, (iii) rapidly deployable in developed and remote coastal locations, (iv) relatively inexpensive, and (v) easy to operate. A compact, inexpensive (< $5,000) 3D terrestrial LiDAR scanner (Blickfeld Cube 1), sampling with a Raspberry Pi 4b+ and powered with a solar rechargeable battery, satisfies all five criteria. Typical ground control point (GCP) -based point cloud rectification methods require excessive gear (e.g., tripods, reflective targets, survey poles, RTK GPS system) that makes rapid, remote deployments cumbersome and time consuming, which inhibits pre-storm deployments. A new point cloud registration approach was applied and compared with a more established GCP-based rectification method. The new registration method only requires an RTK GPS and survey wheel. The iterative closest point (ICP) registration function in MATLAB^Ⓡ was used to register a point cloud measured with a pier-mounted LiDAR at Wrightsville Beach, NC to a survey wheel topographic survey of the beach surface. The LiDAR point cloud was also rectified using a nonlinear least-square-error regression fit to six GCPs deployed in the LiDAR field of view–this method is typically used in camera image georectification. The root mean square error (RMSE) between gridded surface elevations of the registered (GCP-rectified) LiDAR and walking survey point clouds was 0.043 m (0.047 m), with a mean bias of 0.002 m (0.023 m), suggesting the registration approach is a viable alternative.