Also could you please tell me how to make 3D model of the mountain (standing at the bottom) fully? I mean also to include mountain SIDE details (if looking from drone perspective view, trees cover side of the mountain), do I need to set to manual mode and take photos of the side? If so, how to do it, what interval, height and camera angel?
Please have a loot at attached photos, where i demonstrate the example of “lost data” when i take photos only from above and try to make 3D model of the terrain. Here Mountain side view isnt visible to drone from above, because of the hovering trees above terrain.
Thanks for swinging by the DroneDeploy Forum. To clarify, did you post multiple photos? I currently only see one.
Keep me posted,
Here are two more photos, forum was not allowing me to upload more then 1.
By any chance, have you taken a look at our 3D Models support page?
You’re going to need to get a good amount of photos of the side of the mountain and then run it in structure mode when processing on dronedeploy. There isn’t an automated flight mode for this at the moment, but you could set up the camera to take a photo every 2 seconds in dji go 4 and then flight it manually facing the side of the mountain.
Yes I have checked it, but it covers “Easy” places where you can fly in “Circle” trajectory, but my situation is much harder. Sometimes I need to make 3D model of very Narrow corridor of the river. Please have a look at these photos, How is it possible to do 3D model of this corridor? Currently I am working with Hydro Power Plant developing company in Georgia and they asked me to Make 3D model of river corridor (For Intake), but I do not know how its possible with such a narrow area. This is my main goal.
Making Oblique photos is no problem… But it doesn’t cover most of the part because of trees.
Unfortunately mapping this area will be very difficult to almost impossible. 3D modeling is meant for structures, like buildings, houses, or statues - not waterfalls and rivers.
Water and generally anything with a highly reflective surface is problematic for any photogrammetric stitching solution. Photogrammetry works by matching keypoints across multiple photos. It’s the same methodology one would use to solve a jigsaw puzzle. Water presents multiple challenges because it is constantly physically moving slightly between pictures and it is highly reflective. Reflectivity causes specular highlights that also change location between photos. Together, it makes finding keypoints nearly impossible. Therefore, stitching water or any highly reflective surface typically results in spiky, glitchy artifacts in the 3D model.
Hello again @Christina,
First of all, thank you for your reply. As I see you couldnt understand what I wanted to say… I dont need to MAP THE WATER, i want to creat 3D model AROUND THE WATER (ROCKS and GROUND). As you can see from those photos, water occupies very small area, majority are rocks and ground.
Unfortunately this will still not be possible.
I use DD for the image processing, not the navigation. To map a cliff or very steep slope do the following:
a) define the lateral range - say 200m
b) try to identify the slope - i.e. how far away is the top of the cliff compared to the bottom - say 20m
c) figure out the maximum altitude of the cliff - say 60m
now plan your flight under manual control:
d) set the camera to take an image at a regular interval - say 5s
e) set the camera to auto everything
f) pan smoothly across the cliff and find the altitude that gives the best view with the camera pointing down 60-75degrees
g) each pass will advance towards the cliff to keep the framing the same
h) each pass will be at the altitude that gives best vertical overlap parallel to the cliff slope
So I would change altitude by 15m per pass and advance 5m towards the cliff on each pass for a total of four passes. I usually keep the speed under 1m/s but that depends on the distance between the camera and the ground - slower for closer.
If there is dense vegetation there will be no useful data, but for clear rock, even with waterfalls, the images will produce a valid model.
Thanks for sharing this @Arbutus! This is definitely useful info for other people in the community.