Summary of Issue:
Distortion on map
Date Issue Began:
DJI Phantom 3 Advanced
Mobile Device Model and OS version:
Samsung Galaxy S8 latest OS
DroneDeploy App Version:
There has been minimal distortions on my maps before but that’s to expect but in my latest map it has been really bad and i was wondering what the problem is.Here is the picture if you want to see for your self (if you zoom in you start to see it quickly) https://goo.gl/Yf6Ry3
I’m not seeing anything out of the ordinary? Are you talking about the edges of vertical structures and objects? If so that’s to be expected because you are seeing a 2D image of a 3D object that is not perfectly shaped.
Mine are becoming really bad though Michael. I can’t work out what I’m doing differently to what I was doing just a few weeks ago. The accuracy seems to be getting reall6 bad.
I would increase your overlap and make sure you are taking Nadir photos.
Raising the altitude and increasing the overlaps a little will help, but the only way to clean that up is to add some oblique images.
As for the quality “getting worse” it’s not really an apples to apples comparison if it’s different locations and conditions. Maybe go back and reply a previous site that you think looked good as a true comparison?
I fly our office property regularly for testing so I have a constant base.
It’s not the conditions thats causing this for me. Same images @Michael, different results in different processes. About 200-300 oblique used on the above map from me. Took 95% of them out and got great results.
See this post. I can’t explain it. Map details deformation
According to that post you had three completely different conditions created from 3 different sets of images. Even if you fly the exact same flight plan as perfectly as you can in the exact same conditions there will be variations. Same thing in surveying. I can topo the same property twice and will get different results. The GPS of the drone is not accurate enough to expect anything different. With at best 1m accuracy fine details suffer. Your third attempt most closely resembles how I would fly it. There is such a thing as too many images though. All the software besides crosshatch shoots way to many images. You only need about 36 images for a square object equivalent to every 10 degrees. The accuracy doesn’t get much better with more images, but the render gets worse because there become too many tie-points and variables to drape the visual aspect of the image. That all complicated by the density of the angles in your project. The crosshatch is definitely the right solution. I would try to process a map with just those images.
Not quite. I had one single condition and attempted the process three times using a variety of pictures all taken under the same conditions within a 30 minute time frame. I thought it reasonable to give DD the best chance of producing a decent map by handing it a load of oblique photos but as it turns out, on the image above it only want half a dozen from the crosshatch flight rather than the 2-300. Bizarre and not what I’ve previously been advised. In terms of the colourful map in the other thread, it didn’t want any of the Structures images this time but instead preferred all of the crosshatch, the opposite of the successful map in this project. Just doesn’t seem to be any consistency.
But like you said, I’m beginning to appreciate less = more. Nadir and Crosshatch, that’s what I’m going with and pray to Christ I’ve nailed it this time. Thanks for you input. :). J.
Maybe variables is a better word. Best of luck.