I have had absolutely terrific results with the Pix4D photogrammetric software, but have had to develop alternative workflows to using the Pix4D Capture app.
My principal issues with Pix4Dcapture app have been the lack of flexibility and control in planning a mission, and the fact that missions fail when downlink connection is lost, wasting precious battery time.
With Litchi, I can plan a mission at home using GIS software or Google Earth to develop a KML defining the flight line, then import the KML into a Litchi mission in the Mission Planner at flylitchi.com. Vertices on the KML line are imported as waypoints, and mission parameters such as individual waypoint altitudes, flight speed, waypoint turn radii are also set. When the mission is saved, it is automatically synced with the Lichi app on your device, even if you accessed Mission Planner from your PC.
When you fly the mission, you only need to make sure DJI-GO is not running on the device. Once the mission is uploaded to my Phantom3Pro, I set my camera to Interval Capture mode and start shooting when the first waypoint is reached. After the mission has uploaded and begun, the P3P will execute the entire mission autonomously, even if the downlink connection is lost or interrupted (which happens to me regularly).
Using this methodology means that you need to compute (or accurately estimate) your flight line spacing, interval timer capture rate, flight speed and flight altitude for a particular desired ground pixel size and image overlap. Learning to do this will help make you confident and successful. Hint: for the P3P I use a “conservative” 90deg. diagonal field of view; and the image size is 3000x4000, making the image diagonal 5000 pixel units wide. To digitize my grids accurately in ArcGIS, I use the “fishnet” command to build a grid with the desired line spacing, then snap to vertices while digitizing the flight lines.
For 3D object modeling, Litchi offers object-focused flight modes for circling at varying altitudes that look to be extremely effective for that purpose, even though I haven’t done that yet.
Cam,
Do you by chance have any video documentation for this route of planning? I am having troubles creating a 2D grid pattern that will collect images while staying at a set altitude based on the terrain present (my area is hilly). Do you have any suggestions for this? I am really interested in also using the object-focused flight modes for tower inspections, etc.
Best,
Alan
Hi Alan,
After you import the KML with waypoints into a Litchi flight plan you can manually edit the waypoint altitudes to adjust for terrain. You can compute the flight altitude relative to the home-point altitude using a DEM: if you are flying at 100’ AGL to a waypoint 100’ lower, you would adjust the waypoint altitude to 0’.
Because the waypoint altitudes are defined in terms of height above the home point, for hilly or mountainous terrain where you want to plan the waypoint altitudes into the KML exported from the GIS, you can compute the altitude of each waypoint based on an underlying DEM (such as ESRI World Elevation) using the ArcGIS Interpolate Shape tool to convert your 2D waypoints to 3D waypoints with the waypoint Z = DEM elevation. Populate a new field in the waypoint feature class with the DEM Z value using Compute Geometry, then compute the difference in elevation between your expected home point and each waypoint DEM elevation into another new field, then adjust by adding your AGL flight altitude. Then use Adjust 3D Z to make the Z value for each waypoint equal to the computed flight altitude field. Be sure to set up the data frame in ArcGIS that you will export the KML from to use a WGS84 projection and whatever vertical units you will use to specify flight altitude above the home point.
An easier way to deal with hilly terrain is to launch from the highpoint on your site, or adjust your consistent flight altitude so that you fly at that altitude above the high-point on the site, and higher over the rest of the site, letting Pix4D processing work it out.
Cam,
I greatly appreciate the expertise you’re sharing here. I’m a week into Pix4D, can’t stand the flaky connection of my P3S, and have studied the documentation on Litchi. Question on your February post: does each waypoint need to be manually edited? I get the logic, but can’t tell if there is a workflow that converts all waypoints to a given altitude above local terrain. I am also curious if you use “fast” speed, meaning the top speed Pix4D capture uses?
Thanks,
Kelly
I would echo what others are saying here, that pix4d capture lacks the flexibility for planning anything but simple flat ground missions, life in the field is never that simple. The stop start to take each photo is a deal breaker (we are always loosing signal, so cannot use anything but safe mode). Thank you Cam for sharing your process using ArcGIS, will have to try it, we currently use the photogrametrix mission planner solution to map waypoints then upload to Litchi, but being able to use DTM follow of sorts would be ideal. It seems rather extreme to use ArcGIS pro given the tools are there for Pix4D and Esri to build a capture app that actually does that on the fly as part of the Drone2Map package without having to do all those steps individually. We have been trying pix4d on a month by month basis and agree that it does not a bad job of the photogrametry part, but will be using Litchi from now on (downside is we have to be online to upload the mission file which means driving out to service, and back to the job). We are still debating if we should go Agisoft Photoscan or Drone2Map/Pix4D, and without a decent capture app (ideally windows 10 based for our Toughbooks) Agisoft is winning the race.
Say, Cam, any chance you could share .kml, here, or maybe make public on Litchi and post link, would be really appreciative!Please, very interesting stuff!
Here is one of mine, scenic.
Also, FYI HiveMapper.com is rendering 3D maps from MOV files!
They have some work to do, but worth checking out, is free!
Is anyone out there who is interested in flying gently curved, convergent, non-traditional (non-linear/non-parallel) flight paths in order to a) gain a much more diverse view perspective and/or b) help mitigate the well known Structure-from-Motion doming (elevation) error?
Well, here’s an example.
