campatterson
lvl.2
Flight distance : 2014 ft
United States
Offline
|
For those interested in this thread, I will summarize some of my experiences with lidar, photogrammetry, and ground control software for the Phantom3Pro. I fly the P3P to collect orthophotography (terrain corrected, map-accurate nadir aerial photography) for wetland restoration projects where modeling the topography and vegetation is the objective. For some projects, I have airborne lidar, groundscan lidar, and precision GPS surveyed ground control targets all available for the project.
I flight plan a site in ArcGIS or Google Earth, and save the flight-line endpoints to a KML file using a spacing appropriate to the planned altitude, angle of view (91deg diagonal), desired ground pixel height, and length appropriate to my planned flight speed. I import the KML into Litchi Mission Planner online, then save and sync the flight plans (up to four for each of my batteries) to my android device running the Litchi app. I use the editor in the Litchi app to make sure my waypoints are at the desired altitude and other mission settings are correct. When I fly the mission, I set the camera for interval timer mode, typically using 2 or 3 second intervals, and begin collecting imagery when the drone reaches the first way point. I use curved turns to avoid having the drone stop at waypoints. For processing, I use Pix4D ($350/mo rental) and incorporate GPS surveyed ground control. I have found that missions flown at 100' AGL and 14mph with a 2 second interval and 60' flightline spacing yields an excellent orthophoto mosaic at 1" cell resolution or smaller, and the 3D photogrammetric point cloud compares favorably to airborne lidar at 1' DEM cell resolution, and is nearly identical to the ground scan lidar at 1' (while the ground scan lidar resolves the terrain down to less than 1" with mm vertical accuracy within 100' of each scanner setup. the main disadvantage of photogrammetric mass points versus lidar is in dealing with vegetation. I have found the airborne data to be superior for running non-ground point removal algorithms in most cases, but have not yet tried developing ground models from orthoimagery flown beneath the tree canopy. |
|