Point Cloud at strange angles

Hi everyone, after processing the data for a flight in a forested area I noticed that part of the generated point cloud is angled in weird angles. one section is angled one way and another is tilted another.

In the attached pics, the green line is the expected line, while the other colors are all at different angles and not matching the conditions in the field. Our company is still learning and this is a completely new and unexpected result, any input would be appreciated.

Hi demarr,
Can you attach your PDF quality report to this posting? We’ll take a look to see what could be going on.

Clear Creek Flight 2_report.pdf (2.9 MB)

I’ve attached the report pdf.

Hi Blake,
Try changing your Matching back to Aerial Grid or Corridor. The free flight option you selected is probably causing this issue.

Mike, I gave the aerial grid/corridor a shot and the problem persists. I’ve attached the new report from doing step 1. I also noticed that the camera locations (green) also seem the be skewed their geolocated positions (blue) in the image below. Could this be causing the problem?
Clear Creek Flight 2_report.pdf (2.9 MB)

Hi there
You really should use terrain follow in these kind of projects. Seems that there is a quite big elevation difference on the site. Am I wrong? You’ve flown the site with fixed height based on your lift-off location and when there a lot of trees the overlap won’t be sufficient. 6 blocks is never a good sign.

If you can share the dataset I could take a look at it.

Jaakko, you are correct. The overlap is likely the main culprit in this dataset. Using terrain follow would greatly help keep the overlap constant. But it is important to note that this area appears heavily vegetated. Vegetation is always difficult because it has a very complex geometry that isn’t friendly to photogrammetry. Coupled with a high-resolution camera (42mp), this will tend to generate a lot of noise and cause calibration issues. I think it would be worth reprocessing this dataset but try dropping the image scale to 1/4 during step 1. This will tend to negate the effects of the vegetation. It might help.

For future flights in heavily wooded areas would you recommend having less overlap in the photos? are there any other best practices for processing data from wooded/heavily wooded areas? or a link to the resources.

I gave the 1/4 image scale a run and it seemed to mostly fix the issue, however a section remains at an angle. Fortunately the section at an angle isn’t very relevant to the rest of the site.
Clear Creek Flight 2_report.pdf (2.5 MB)

The key to heavily wooded areas is high overlap. You’ll want to be in the 80-90% range. Don’t exceed 90%, though.

Did you try with the Aerial Grid or Corridor matching option?