Accuracy in Orthomosaic Output is not coming

[Andysmith321]

What measures one should take to achieve less than 10 cm horizontal and 20 cm vertical accuracy for Orthomosaic. I am working on a project (Agriculture Land Survey) and the client's requirement is accuracy less than 10 cm. We are using images from Talon with PPK. Currently we are following below work-flow.

Work Flow -
1) We have done PPK Processing (All fix no float value). Then geotagging of images.
2) Adding Geotagged Images in Chunk 1
3) Photo Alignment (High)
4) Added Markers, Completed Marking and Optimised images.
5) After Optimisation Error (Pix) was around 9. (How to better this value?)
6) Dense Point Cloud Generation (High)
7) DEM
Build Ortho

I feel the Pix Error is the culprit, how can it be rectified. Is my process is correct?

Request you all to please guide me to achieve Horizontal and Vertical Accuracy of Orthomosaic.

[DayGeckoArt]

Are you using ground control points? And are you sure the coordinates of the GCPs are accurate?

[Andysmith321]

Are you using ground control points? And are you sure the coordinates of the GCPs are accurate?

yes we are using GCP's and coordinates are correct. I feel Geotagging is an issue as the "Pix Error" after optimisation is High, It is coming around 9. I feel if we can better this Pix Error then we shall get desired result. But I don't know how to do it?  Can someone help?

[Paulo]

Hi AndySmith,

your PPK determined positions how are accurate are they estimated and what is the accuracy do you have on these? If you have a doubt on this accuracy then you can just put a 10 m accuracy vs example 1-2 cm on GCPs. And during optimisation the constraint on cam positions will be loose and on GCPs strong so letting all cameras positions to move to fit the GCPS. Or just untick them so camera positions will not be taken into account. Also be sure the datum is same for both cameras and GCPs. Often the vertical datum is different, ie. camera with elipsoidal h (WGS84) while GCPS are referred to a MSL ortho heights and the difference can be considerable...

Maybe this can give you some avenues for your problem,,,,

In general with good camera, control, you can get easily a product with 1 GSD or better accuracy in XY and 2-3 GSD in H or flying at 7-8 cm GSD should be able to satisfy your client needs..

PS. If you can share a small dataset, I could look at it and give you my best recommendations,

[Kiesel]

Hi Paulo,

...or flying at 7-8 cm GSD...

is that a typo?   

Best regards,

Kiesel

[Paulo]

Hi Kiesel,

no i do not think so. Flying at 7 to 8 cm GSD with good control would certainly satisfy the 10 cm planimetric accuracy as well as the 20 cm vertical accuracy required for ortho...

What is your opinion?

[Kiesel]

Hi Paulo,

it was a misunderstanding at my side. You meant that you fly at a height to get 7-8cm GSD and not that you fly at 7-8 cm as I have read it, right? 

Best regards,

Kiesel

[Andysmith321]

Dear Paulo,

Thanks for the reply.

Additionally after going through several posts, I have used "Gradual Selection Tool" for reducing the error. In order of reconstruction uncertainty, projection accuracy and reprojection error. I was able to achieve the desired value of 2. However in the process too many points were delete. MY project is of 5 GSD, I wonder if this plays any part in determining the threshold value for these 3 parameters of gradual tool.

Can someone help how to determine the threshold value for reconstruction uncertainty, projection accuracy and reprojection error? And does GSD plays a role in that?

Regards,
Andy

[Andysmith321]

Dear Paulo,

Thanks for the reply.

Additionally after going through several posts, I have used "Gradual Selection Tool" for reducing the error. In order of reconstruction uncertainty, projection accuracy and reprojection error. I was able to achieve the desired value of 2. However in the process too many points were delete. MY project is of 5 GSD, I wonder if this plays any part in determining the threshold value for these 3 parameters of gradual tool.

Can someone help how to determine the threshold value for reconstruction uncertainty, projection accuracy and reprojection error? And does GSD plays a role in that?

Regards,
Andy

Any update on this?

[Andysmith321]

Dear Paulo,

Thanks for the reply.

Additionally after going through several posts, I have used "Gradual Selection Tool" for reducing the error. In order of reconstruction uncertainty, projection accuracy and reprojection error. I was able to achieve the desired value of 2. However in the process too many points were delete. MY project is of 5 GSD, I wonder if this plays any part in determining the threshold value for these 3 parameters of gradual tool.

Can someone help how to determine the threshold value for reconstruction uncertainty, projection accuracy and reprojection error? And does GSD plays a role in that?

Regards,
Andy

Any update on this?

Still looking for an answer, can someone help me on this?

[Kiesel]

Hello Andysmith321,

I have used "Gradual Selection Tool" for reducing the error. In order of reconstruction uncertainty, projection accuracy and reprojection error. I was able to achieve the desired value of 2. However in the process too many points were delete

I use specific values for each project, so that no more than 40-50% points are deleted for each step (reconstruction uncertainty, projection accuracy, reprojection error).

Best regards,

Kiesel