A totally wrong survey

[PeteM]

I use the Phatom 4 for topographic surveys for several months.

My last flight I did was return an area with elevation values TOTALLY wrong (climbs instead of downhill...)

The strange thing is that this only happened in a little land area I've mapped.

This is something I can not explain.

The drone flew to 60m, with 80% overlap, and the day was a little windy. I would add that the surface was almost flat and of uniform white color.

What happened ?

 

[Richard R]

you don't mention which stitching software you were using?

 

[PeteM]

Photoscan Pro

 

[Richard R]

Don't know about PSP, I've used both Maps Made Easy and Drone Deploy with good results. Any photo stitching program can have problems with flat, uniformly textured surfaces. There just isn't enough uniqueness for the program to do the calculations properly. Sentences wasn't clear, but I interpreted what you said in your original post to mean that you only had problems with a small portion of the total map. Did the photomap of that same area come out ok? If so, you might want to check the exif files for some of the images over that portion to make sure that something wasn't corrupted in it. Was this an area that you had mapped before or totally new mission?

 

[Meta4]

The drone flew to 60m, with 80% overlap, and the day was a little windy. I would add that the surface was almost flat and of uniform white color.

Was the land surface so uniformly white that the software couldn't find points to match properly?

 

[PeteM]

The entire area to be mapped is around 20,000 sq. M. About 60% of this area is made up of virgin soil, with very low vegetation and some scattered trees. The remaining 40% of this area (in the center) is a kind of hole filled (up to the level of the surrounding soil) from a material very similar to a white sand. This "white material" forms a sort of dome surface, with the highest part in the center and the lower parts bordering the virgin soil.

The 3D model is correct from the top of the map up to the middle (the top of the "dome of white material"), from then on instead of lower elevations I find higher elevations.

On the white surface there are obvious signs of truck passage and there are many small "black spots" so I think the common points can be found by the software. The surface is almost flat with a height difference of not more than 2 m distributed over a large area.

The orthophoto is perfect. The distances are perfect. It's an elevation problem.

 

[doktorinjh]

I've seen PS get confused as to what's up and down, so try flipping the model to see if that could be an issue. I assume that you're using exif data from your P4? Make sure that everything is importing into the correct lat and long coordinate columns. You didn't mention if you're using GCPs, but also double check that they're in the correct columns too, if applicable. Please let us know how it goes.

 

[Richard R]

Double check the elevation in the exif file for the images of the area where you are having the problem. A while back, DJI switched to recording the GPS derived MSL altitude rather than the barometer derived AGL height. if there was a glitch in the GPS over that portion, the estimated heights could be off affecting the resultant elevations. DJI still has the elevation data in the XMP portion of the exif file, but some readers don't get to that data.

 

[r62ewa]

My altitudes in the EXIF are 99% of the time read by Pix4D as negative as the AltitudeREF is set to BelowSeaLevel rather than AboveSeaLevel. Very frustrating to see the EXIF read 39m altitude and then Pix4D interpret it as -39m altitude.

 

[Richard R]

And that would surely screw up your elevation maps. Did you try loading the images into another program such as Maps Made Easy? Not sure if your problem is related to the same glitch that I had. I have 3 P3As all running FW 1.9.6. All 3 gave me good orthomap results. But on 2 of the 3, both Maps made Easy and Drone Deploy gave me good elevation while the 3rd always gave me incorrect values. They would show my bird flying only 1/3 to 1/2 as high AGL as it actually did and elevations would be compressed to the same levels! Tried everything; flying crosswind, slowing down, increasing overlap - nothing worked. And checking the exif and XMP files, everything look ok. Then I had a 'brain hiccup'. I got to thinking that for the 2 birds yielding good elevations, I had actually upgraded the firmware while the 3rd came with that version already loaded. Couldn't see where I had anything to loose, so I flash upgraded the same version from the bin file that I had on my computer. Guess what? My elevation maps look just as good with this bird as with the other 2! I know that there is no logical explanation for it, but there was something not being set right in the exif file with the FW txt had been loaded and it apparently was fixed when I reloaded the FW. If it happened to me, something like that might be going on with your FW.

 

[r62ewa]

So when you view your EXIF data of a photo direct from your Phantom the AltitudeRef is not stored as Below sea level as mine is which you can see from the image below? When Pix4D reads the EXIF it interprets this a negative elevation. Are yours stored as Above sea level when you view a photo direct from your SD card?

 

[Richard R]

Definitely 'Above Sea Level' as shown below.

 

[r62ewa]

Huh, why is it that mine is setting it as Below sea level? Is there a workflow to have photogrammetry software process using the Relative Altitude stored in the XMP as is show in your screenshot?

 

[Richard R]

Recently asked the MME people about this. Their response was that they use the first photo in a set which is taken while the bird is still on the ground as a geo reference then use the relative altitude in the XMP portion of the files to fix the height agl. As I mentioned earlier, after doing the reload, all 3 of my birds seem to yield good elevation maps. Still can't answer why you are getting 'Below Sea Level'. Any idea what the altitude really is?

 

[r62ewa]

The Relative Altitude seems to be correct. I too use Map Pilot and I used an EXIF reader and looked at the first image taken and it has a Relative Altitude of 0m (the geo reference image taken at ground level) and the flight of my grid and subsequent pictures seems to be correct at 54m. So the altitude stored in the EXIF appears to be correct, it just stores it as Below sea level instead of Above for some reason. I just recall Pix4D did give me an explanation of why my Phantom is doing this but that doesn't explain why mine does it and yours doesn't. I would like Above as well for PIx4D processing.

 

[Richard R]

With my glitch, no one (MME, DD even some of the others on the forum) could find anything wrong with my exif files. But I always got that error in my elevation maps. Again, I have no explanation for why it worked, and make no claim to know if it could straighten out your issue, but can you try to reload the FW on your bird?

 

[r62ewa]

That may be something I would have to try. I have not done it before so will have to do a bit of research on here on how to accomplish it. I always found something strange about my Phantom 4 as I often can't get full grid image capture using any of the Apps. I usually get the first 2 lines of a grid and then sporatic captures for the remainder of the grid. Doesn't seem like many people have this issue but I do for all capture Apps. This along with the Below sea level is frustrating.

 

[Richard R]

Believe me, I know full well how frustrated you must be. I probably flew a dozen missions trying to figure out why my elevations were wrong.

 

[r62ewa]

Are you using GCPs as well?

 

[Richard R]

Not yet. At this time, Im more interested in the relative accuracy (distances between points within the map) than I am with the absolute. Even the absolute is sub meter when compared with Google map!