Processing Large PPK datasets

northsurvey · October 6, 2020, 5:59pm

Hi Community,

We are trying to process 1800 PPK images with 37MB per image over 350 acres of a gravel pit with some areas of dense forest. We were only able to acquire 65% front overlap and 65% side overlap with the Sony RX1RII on the Quantum Systems Trinity at 120m AGL because of trigger speed. Which could be the source of our issues. However we have divided the project up into 3 and have had success on the initial process. Ideally we could be able to process the whole project as one. Here is what we have attempted with the full dataset.

Processing the full dataset has failed 3 times.

1st Attempt:
Image Scale : 1/8
Calibration method : standard
Internal Camera Optimization : All
NO GCP

Results:
80% images calibrated
Significant bend

2nd Attempt:

Image Scale : 1/2
Calibration method : Alternate
Internal Camera Optimization : All prior
4 GCP

Results:
90% images calibrated
slightly less bent result
significant error with check points

3rd attempt:
Custom keypoints: 25,000
Image Scale : 1/4
Calibration method : Alternate
Internal Camera Optimization : All prior
30 GCP

Results:
FAILED PROCESSING

Any advice would be appreciated.

Thanks!

Blaz · October 7, 2020, 1:37pm

Hi,

Thank you for the detailed description of the workflow you follow.

As you already mentioned, the fact that there was only a 65%/65% overlap between images and that there is dense vegetation in the area could present an issue for processing.

Could you share the quality report and a few screenshots of the processed project to get a better overview of the scene?

In general, the processing options that you tried to adjust can help with the project calibration, and we recommend the same adjustment:

Process with lower Keypoints Image Scale: This processing option can lead to a higher number of calibrated images than the default original keypoint image scale. For more information: Menu Process > Processing Options… > 1. Initial Processing > General.
Increase overlap (>80%).
Flying at a higher altitude. This tends to improve results in forests and dense vegetation environments. For more information about the flight plan: Step 1. Before Starting a Project > 1. Designing the Image Acquisition Plan > a. Selecting the Image Acquisition Plan Type.

Do the images you use have accurate IMU angles? If yes, then you could try processing with the Accurate geolocation and orientation pipeline.

Best.

northsurvey · October 7, 2020, 8:11pm

Hi Blaz

thank you for the help, attached are screenshots of the quality report as well a couple overviews of the scene.

unfortunately the images do have accurate geolocations, but lack an accurate orientation angles. Therefore we have not been able to us the Accurate geolocation and orientation setting.

please refer to the link below for the screenshots
https://northconstructionorth-my.sharepoint.com/:f:/g/personal/gavin_north-construction_com/Ejggjoqv_wpBiP-tqWb2BK0B28np9tUiTaYZwQG0AvWuyQ?e=mPQQBy

Blaz · October 12, 2020, 2:26pm

Thank you for the screenshots.

From what I could see you did a very good job by making sure that there is enough overlap between the subflights. All the subflight processed together in one block.

The uncalibrated images can mostly be seen at the edge of the project and in the area close to the center which, most probably, contains dense vegetation or other features that present a challenge for image-based processing software.

I think that you could get better results by using a higher overlap between the images and by flying higher. Areas with dense vegetation have a bigger chance to get calibrated if a bigger area is visible on original images.

Besides adjusting the processing options mentioned above, you could also try adding additional MTP to the project to assist the calibration.

Hope this helps.
Best,