Subscribe to this thread
Home - General / All posts - Geotif with white holes
ponsrud11 post(s)
#09-Apr-24 11:56

Hi, so i recently downloaded manifold and have started to get used to how it works. However, I have a problem when reading geoTIF files. For some reason I get these white holes spread over the image and I dont know why. When I read the files in other software there is no problem, but here it clearly is. Do any of you know what can cause this issue? See image below for example

dale

637 post(s)
online
#09-Apr-24 20:41

That look like null data. Can you share the image or provide the source?

ponsrud11 post(s)
#10-Apr-24 06:54

Hi, thanks for reply.

I dont know what the problem is, but it is only when reading in manifold that the pixels get blank. Works fine in other software.

It happens with all my tif files, so its a problem for me since I want to use manifold to convert the TIF to ECW files.

Hers a link do a tif ifile where the error occurs: Manifold_Forum_example

(Let me know if the link dosent work!)

Bernd Raab66 post(s)
#10-Apr-24 07:07

Hi,

I opened your data in GlobalMapper and in QGIS and in Manifold 8, every software shows nearly the same "holes". So it is an issue with the data.

ponsrud11 post(s)
#10-Apr-24 07:28

Wow ok, thats wierd, I tried the same thing in QGIS and you are right. But when i open it in FME or windows photo, there are no holes:

Is it something in the way the different softwares represent null data?

Attachments:
no_holes.png

dale

637 post(s)
online
#10-Apr-24 07:40

Agree Bernd, I get the same results. Tried QGIS, M9, and PS.

The linked image is derived from point cloud data, so structure from motion. There are nulls and other visual artefacts, mostly within areas of low contrast. They are errors in either processing, or in drone mission planning (or both)

There are other errors visible in the sample image with areas with low match points. In areas of low visual contrast, there are obvious processing artefacts.

I checked your image against Google, and Bing image servers, the image is not quite registered correctly, with a west shift of tens of metres.

ponsrud11 post(s)
#10-Apr-24 07:52

OK thanks.

But do you have any answers of why there are some software that manage to fill the gaps? I am using DJI terra to process the data, and there are no holes there either.

The image is taken from drone, so there are indeed motion involved, do you think higher overlap would give a better result?

Bernd Raab66 post(s)
#10-Apr-24 08:33

In Manifold you can try to fill the gaps.

See video example: https://www.youtube.com/watch?v=YIhndI_FIXM

Trying to open your image in ArcGis gives the same result (holes). So GIS-Programs handels pixel-values different than image processing programs.

ponsrud11 post(s)
#10-Apr-24 08:47

Yes, it seems like there are som differencies in how the software read pixel-values. But ok, then I will try with higher overlap. Thank you for helping me figure it out:)

I have tried the fill function and it is sort of doing the job, but the result file gets this blue overlay all over eventough I have tried to change the StylePixel:

Attachments:
blue_overlay.png

rk
627 post(s)
#10-Apr-24 10:34

The blue overlay looks like transform preview. Are you sure it is not that?

ponsrud11 post(s)
#10-Apr-24 11:07

Hi, yes thats absolute true, my bad! But the problem is still there after i press the transform button as well. The result file is then a black/white image, and if I change the StylePixel to the same as the original I get the red result. See attachments

Attachments:
black_white.png
red_overlay.png

rk
627 post(s)
#10-Apr-24 11:39

Ah, filling/interpolation works per channel. The result is single-channel image. You could do 3 interpolations and merge channels back. Not super friendly, but doable.

ponsrud11 post(s)
#10-Apr-24 12:36

OK, thank you. I shall try it out :)

rk
627 post(s)
#10-Apr-24 14:15

It seems that whatever software exported the TIFF marked total black as NODATA value. You should probably replace NODATA/Missing Pixels with just RGBA=0,0,0,255.

No interpolation needed.

gdalinfo output:

Band 1 Block=256x256 Type=Byte, ColorInterp=Red

NoData Value=0

Unit Type: metre

Band 2 Block=256x256 Type=Byte, ColorInterp=Green

NoData Value=0

Unit Type: metre

Band 3 Block=256x256 Type=Byte, ColorInterp=Blue

NoData Value=0

Unit Type: metre

Band 4 Block=256x256 Type=Byte, ColorInterp=Alpha

NoData Value=0

Unit Type: metre

s

ponsrud11 post(s)
#10-Apr-24 18:54

Ah ok, I guess I can do the replacement in the transform pane?

dale

637 post(s)
online
#10-Apr-24 12:37

I have some experience in DJI Terra, more with PIX4D.

The rough rule of thumb for areas of high heterogeneity so snow cover, sand, canopy etc, fly a higher (altitude) mission, and reduce the resolution.

Are you using a DJI lidar like the L2?

I use an M300 with the Zenmuse L2, in areas of sand/beach, as well as dense tree canopy. Running higher overlap will assist in obtaining more match points over those areas. Your snow covered areas, and the bare tress are areas where the existing imagery is not as suitable.

If using LIDAR in Terra, export the data as LAS, then bring that into M9.

Another option is to bring your imagery in PIX4D and process it there. That may do a much better job than Terra.

<edit> link to helpful video on registering drone imagery

ponsrud11 post(s)
#10-Apr-24 18:59

Yes, I am using the L1 scanner. This is the first time I have launched my images in a GIS software, so I havent been aware of the white holes, but I shall experiment a bit with higher overlap and see if it helps. Pr today i am usualy using 60% overlap. What do you usually use?

Do you use M9 to enhance the laserdata?

dale

637 post(s)
online
#10-Apr-24 22:21

M9 is a fantastic tool for both imagery and pointcloud data (LAS) Have read of the manual for LAS functionality. There is a bunch of videos on the Manifold channel, good place to go as well.

In terms of improving the imagery collection, and noting RKs null data post above, try increasing the overlap. Survey a subset of your data collection area, over the most problematic area. See how that works.

Manifold User Community Use Agreement Copyright (C) 2007-2021 Manifold Software Limited. All rights reserved.