Need help with creating an orthomosaic

Spades0 · Mar 5, 2018

Hi

I was recently given the task to create a orthomosaic using our Phantom 4 Pro of a rock outcrop in South Africa for a geological research project. In preparation for this task I downloaded the Precision Flight app and went for a practice session the weekend before. After the flight I uploaded the photographs onto the PrecisionMapper website and got a perfectly beautiful result (see attached image called "Church"). Thereafter I was confident to try and go and map the rock exposure.

However, the map of the rock exposure was a complete failure (the attached image is called Tennisball Marker). There appears to be a lot of black "fractures" in the final product and the result doesn't make any sense at all. When I looked at the pictures the drone took they also appeared to be a bit erratic; in some cases photographs do not seem to overlap as they should and it seems like the drone changed its direction in the middle of a traverse when it was supposed to fly in a straight line on one occasion.

I have no idea what went wrong. Between the successful and failed attempts I updated the firmware of the drone. Could this possibly have had an effect? The drone also flew at a different altitude (45 meters for the failed attempt, 100 meters for the successful one). The elevation at the failed site was also much more irregular compared to the successful site which was relatively flat.

Any help would be appreciated! I will need to go remap the rock exposure again sometime in the future but before I go I need to know what went wrong the first time.

Thanks!

Ataim · Mar 5, 2018

Not sure on the software.. I use Pix4D. But a few questions...

If it flown at 100 but was expecting 40 that would be my first thing to look at. Big elevation changes, software might have problems. How fast did you fly? What angle or angles was the camera set at? What was the camera set at (quality), is the memory chip fast enough? was it windy? What program did you use to fly with? What is quality of intended final deliverable? Survey grade? If so did you have any ground control points, if so how many?

drvrbobob · Mar 5, 2018

I am new at this but could the black be shadows, seems they are all oriented the same way.

Meta4 · Mar 5, 2018

Spades0 said:
I have no idea what went wrong. Between the successful and failed attempts I updated the firmware of the drone. Could this possibly have had an effect? The drone also flew at a different altitude (45 meters for the failed attempt, 100 meters for the successful one).

Firmware? Probably not a factor
Altitude? Maybe
From what I see in the pic, it appears that there's a lot of indistinct detail (like the top of a forest).
That makes it hard for the software to lock on to definite points to match between images.
Mapping areas difficult sites like forest cover, or similar subjects requires more altitude and a bigger overlap.

Spades0 · Mar 7, 2018

Ataim said:
Not sure on the software.. I use Pix4D. But a few questions...

If it flown at 100 but was expecting 40 that would be my first thing to look at. Big elevation changes, software might have problems. How fast did you fly? What angle or angles was the camera set at? What was the camera set at (quality), is the memory chip fast enough? was it windy? What program did you use to fly with? What is quality of intended final deliverable? Survey grade? If so did you have any ground control points, if so how many?

Hi Ataim, thanks for the response.

The first practice flight was at 100 meters while I set the altitude for the second (failed) attempt at 45, so it did not behave in an unexpected manner. I flew at a speed of about 6-7 m/s and the camera was pointed directly downwards. I haven't found a setting to change the quality of pictures taken with the camera with the app, but I assume it is the highest possible quality the camera can deliver (4864 x 3648 resolution). I am not sure if the memory chip is fast enough. I guess it is possible that this contributed to my problem. It was a bit windy, not too bad I think. To fly I used the PrecisionFlight app. I was hoping to get a final resolution of about 1 pixel/cm. I did not have any ground control points; what exactly do you mean by this?

Spades0 · Mar 7, 2018

drvrbobob said:
I am new at this but could the black be shadows, seems they are all oriented the same way.

Hi drvrbobob,

I do not think so, I don't see any shadows like that in any of my photographs.

Spades0 · Mar 7, 2018

Meta4 said:
Firmware? Probably not a factor
Altitude? Maybe
From what I see in the pic, it appears that there's a lot of indistinct detail (like the top of a forest).
That makes it hard for the software to lock on to definite points to match between images.
Mapping areas difficult sites like forest cover, or similar subjects requires more altitude and a bigger overlap.

Thanks Meta4, I will keep that in mind. I am starting to suspect that the low altitude together with a too high flight speed might have been a problem. Unfortunately the app I'm using (PrecisionFlight) does not seem to give me the option of selecting a flight speed. I think I'll try a different app next time.

Meta4 · Mar 7, 2018

Spades0 said:
Thanks Meta4, I will keep that in mind. I am starting to suspect that the low altitude together with a too high flight speed might have been a problem. Unfortunately the app I'm using (PrecisionFlight) does not seem to give me the option of selecting a flight speed. I think I'll try a different app next time.

If it's like DroneDeploy, it will calculate a flight speed based on the overlap and altitude you set.

doktorinjh · Mar 7, 2018

Spades0 said:
the rock exposure was a complete failure

If you can post the images, I can run it through Pix or PhotoScan and generate a coverage map, if PrecisionFlight doesn't do that for you. That will show whether or not it was the altitude or the homogeneous terrain that could be the cause.

Spades0 · Mar 8, 2018

doktorinjh said:
If you can post the images, I can run it through Pix or PhotoScan and generate a coverage map, if PrecisionFlight doesn't do that for you. That will show whether or not it was the altitude or the homogeneous terrain that could be the cause.

Hi Doktorinjh,

I would be happy if you can take a look at it. However, the file is too large to attach here. Can I send it to you in some other way? (Google Drive maybe?)