DSM - DTM showing negative values

My DSM and DTM have about 11 cm of GSD. DTM = 5 x GSD (2.24 …).
To find plant height I do DSM-DTM but the resulting raster containing plants height has negative values in various places.
How can I correct this problem and increase the accuracy of pixel height values?
I have used the parameters below … Should I change them?

Hi,

Are you using any type of RTK/PPK drone? I am having some doubts about the accuracy of the calculation DSM-DTM, because you also need to account for the XYZ uncertainties, especially when processing images captured with consumer drones (the GPS accuracy can reach a few meters). One important replacement for RTK/PPK accuracy would be the homogeneous distribution of GCPs across the image dataset. This can increase the XYZ accuracy and provide more reliable values. Please take a look at the following article about the accuracy that users should expect in Pix4Dmapper: Accuracy of Pix4D outputs.

Please also note that the DSM and DTM are generated from the filtered point cloud. The filtered point cloud is generated from the densified point cloud that is generated based on the Automatic Tie Points. I would closely check the quality of the DSM and DTM, because problematic areas where the DSM and DTM have big triangles or are badly reconstructed are areas where the densified and filtered point cloud have bad, few or no information at all. These reconstruction problems can also affect the DSM-DTM calculation.

Cheers,
Teodora

Thank you Teodora,
I used a GNSS RTK to mark GCPs in the area of interest…
See the Quality Report attached, please. What’s wrong ?

CY_27032019_Multi_report.pdf (2.4 MB)

Dear Teodora and Support members,
Interestingly, according to field measurements, plants have an average height of 80 cm, but the Pix4D volume tool is also providing negative volume values …
What should I do to repair this? Should I remove the filters from DSM? Which are ?
I ask for help to you.
Is there any more accurate tool or software on this issue?

Hello again,
From the quality report, it seems that the DSM has a resolution of 2.25 cm/pixel, while the DTM has a resolution of 11 cm/pixel.

According to the rules of digital image processing, operations between images can only be performed if they have the same pixel size. The easiest way would be to resample the DSM to the spatial resolution of the DTM (11 cm) and then perform the computation.

Cheers,
Teodora

Ok Teodora,

I had already resampled the DSM pixel to 11 cm in ArcGIS and honestly I could not understand the reasons for the negative values in the DSM-DTM operation. I have been able to solve this issue by using plant height calculations with Global Mapper 20, but individually. I think we should wait for future adjustments to the DTM calculation in the next versions of Pix4D.

Thanks,

Walter

PS: If there is a solution for the case, please post it here in the community ok?

Hi Walter,

Did the DSM and DTM generated from Pix4Dmapper have correct elevation values? To be honest, this is the first time I hear about calculating the plant height using elevation or surface models generated from Pix4Dmapper. How did you measure them in Global Mapper?

Cheers,
Teodora

Hi Teodora,

GM has a distinct way of calculating the height of objects… It uses only the DSM and takes as reference the nearest soil surrounding the plants. It is quite accurate when compared to field measurements.

Cheers,

Walter

Hi,

The DSM and DTM generated from Pix4Dmapper should be correct, but measuring plant height is tricky because of the interpolation algorithms used by Pix4Dmapper. Also, don’t forget that there is a certain incertainty in the x, y, z directions. From a research perspective, I would rethink the strategy. However, I have no better suggestion at the moment. Hopefully, you’ll get feedback from some users who have similar research questions.
Cheers,
Teodora

Hi Walter,

Can you please describe how do calculated heights using GM? I have the same issues you mentioned here and wanted to try GM if it will give me better results.

Cheers,
Jayfred

Dear,
you need to load a digital elevation model (DSM) and orthomosaic in the GM, in that order.
Then, you can use a shapefile or the profile tool to demarcate your area of interest on the orthomosaic
and thus calculate the volume or height of the plants in this area.
This tool supports several types of configuration.

Hi
I’m facing the same problem.
Does anyone has an idea to fix this?

Hi Thomas,

Can you please elaborate a bit more on this? Is the issue related to having negative values for the plant height or are the DSM and DTM incorrectly generated? A quality report would be very helpful.

Cheers,
Teodora

Hello! I have a problem with the DSM generated in an area where the elevation is approximately 16 meters above sea level, it returns only negative values of the order of -50 meters. What could be the problem?

Hi nicolaspaz,
Are you using GCPs in your project or do you have an RTK enabled drone? If not, then you are relying on the onboard GNSS of the drone to establish elevations which can be very inaccurate. I suggest using GCPs as it will correct for this.

Hello! i have the same problem again.
In another place with a similar (but not the same) drone DJI Phantom 4 pro. when i get the DTM the elevations are in negative values. How can i fix it? Thanks!

Hi nicolaspaz,

Are you near sea level? If so, what vertical coordinate system are you using?

Approximately 10 meters above sea level. Output Coordinate System WGS84 /UTMzone 20S(EGM96Geoid). The values I get in the DSM range from -97 in the lowest areas to -93 in the highest

Hi nicolaspaz,
This is somewhat expected as the EGM96 is a global model and may not fit the earth perfectly everywhere. It can easily be 10 meters off. I would suggest using a geoid that is local to your area.

Thanks for the reply. The images were already taken in that system, is there a way to change it in the processing with the PIX4D? To have a consistent output? Cheers