Antenna Tip Not Reconstructed Despite Valid MTP

PIX4Dmatic
1

Dear Pix4D Support Team,

First of all thank you in giving solutions. I am reporting a reconstruction issue in Pix4Dmatic v2.6.0 (Windows 11) involving a thin vertical structure (rooftop antenna).

We captured a construction site with a rooftop antenna using multiple circular orbit passes and nadir images. We marked the antenna tip as a Manual Tie Point (MTP) in 11 images with a reprojection error of 1.4 px. Despite this, approximately 1 meter of the antenna tip is missing in the dense point cloud.

The MTP was correctly placed and accepted by the software, yet the reconstruction still fails to represent the tip of the antenna. Our client requires the exact height of this antenna, and the current output is insufficient.

How to counter this issue, is there any mistake in our workflow? Or this is the limitation of the software? Up to date we understood the MTP workflow (densyfing specific area or details) exactly for this matter. Or we did we understand the concept of MTP wrong?

Screenshots of the issue are attached for reference.

Thank you for your time — we look forward to hearing from you.

Best regards,
Humayun Sajjad
Octacon Geotechnik GmbH

Screenshot 2026-06-16 123855
Screenshot 2026-06-16 1238552521×1235 598 KB

2

Hi humayunsajjad.official,

Thin structures have always been very challenging to reconstruct using traditional photogrammetry. Have you tried the Gaussian splatting option? This tends to reconstruct thin structures extremely well. I suggest giving this a try.

3

Thanks for the suggestion — we did in fact already try 3DGS as well. However, we find the question a bit puzzling: if there is no dense point cloud to begin with, no seed can be generated for a 3DGS reconstruction either. In other words, the antenna is missing from the 3DGS result just as much as from the point cloud — so this approach doesn’t really address our issue.

Our original question was whether an MTP (manual tie point) would cause the point cloud to be refined/densified in that area, and why this isn’t working for the antenna specifically.

The follow-up question is: how can I get this antenna reconstructed properly, and how many MTPs would I need to set if a single one isn’t sufficient?

Thank you for your attention.
Regards,
Humayun Sajjad

4

The MTP will likely not generate more points. My suggestion is to increase the overlap settings. There needs to be more common keypoints between overlapping images.