Support Website Contact Support Blog

Merge / W0009

Hi Christina

O am new to pix4d and learning the system with a small project of a deck area. The issue i am having is that three sides of the deck i.e. walls appeared in the point cloud and mesh but the 4th side i.e. fence did not.
As a check I then created another project with the same photographs just for the fence and they were modelled. Therefore I can’t understand why they did not model in the 1st project.

I then decided to merge both project but I keep getting the W0009 warning coordinate output values are not the same. I am not using any GPC - only MTP. However as it is only a deck area and I am using a DJI phantom 4 pro I dont believe I need a GPC .
My final question is: if a GPC is required how would I attain that as I don’t have any of the control matts or the RTK software etc. Is there another way to establish a GPC?

Hello and welcome to our Community!

Please follow the instructions of the following article for merging:

(MTPs are necessary for merging)

Some times when the overlap between different parts of our models is not enough, some parts may get discarded when processing all images together. Working on individual projects and then merging with the instructions of the above article may give a complete output.

Regarding the warning you get, please check that the output coordinate system of all subprojects is the same:

I hope it helps :slight_smile:

Hi Christine

I did try as suggested, I included the same MTPs for both projects as per the video and to ensure overlap. I checked the coordinate output and made sure it was the same. However when I merged the projects the fence portion didn’t show and was listed in the photographs section as uncalibrated…Yet when I created the project with the fence photographs by themselves the fence shows up clearly and the photographs are listed as calibrated

50 photographs taken on the same. flight path - approx 4 minutes on a back deck of a building in NYC to test and learn about the software. The company wants to use tis software for major building projects in NYC but this is somewhat disturbing if 50 photographs cant produce a full representation of the perimeter of the deck. Can you help out with this?


Could you please send us the quality reports of the merged and the individual projects, so that we can better understand the case? The quality report can be found here:

: …\Project_name\1_initial\report\project_name_report.pdf

Thank you!

Hi Christine

I am attaching the quality reports in PDF but in summary the following is happening.

POINT 1. Thus is the First project F1 - which included all the photographs taken. As can be seen the right side (WOOD FENCE) did not populate yet the associated photographs are indicated as uncalibrated. Another issue is the portions of the metal fencing on the left wall did not complete either. So I then decided to perform a project just for the fence to determine if the photographs were the issue. See Point 2.

POINT 2. As can be seen tall but three of the photographs for the WOOD FENCE populated by themselves - this is what should be included on the right side. I then did a project without the WOOD FENCE (FENCE TEST) and merged that with the WOOD FENCE PROJECT - SEE BELOW POINT 3.

POINT 3. However when both projects were merged (FENCE TEST) once again the WOOD FENCE did not poplulate and the photographs for the WOOD FENCE are listed as uncalibrated

F1_report.pdf (182 KB)

fence test_report.pdf (170 KB)

wood fence_report.pdf (127 KB)


Thank you for your detailed description.

In general, you followed the right workflow. You extracted the problematic part and processed it separately. However, I would say that in your project the problem is the acquisition plan. The object you are trying to model is very uniform, homogenous (concrete, metallic, wooden elements) which impedes software algorithms to work efficiently. In my opinion, the solution would be to acquire the dataset with higher overlap and process it in one go.


Thanks Christine,

I will certainly attempt that. However I cant understand why the fence is modelled perfectly when done as a project alone. To me the algorithms are working in that scenario so why wont t work in a combined manner?


Apparently, in your project algorithms don’t see your dataset as a holistic structure. Therefore, the logic of the software breaks at some point. The higher overlap between images should help him understand that the content of the images varies, yet it’s the same object, just with a different texture.