Support Website Contact Support Blog

Bowl effect

 

Hello,

I am trying to detect the ‘bowl effect’ on my results. Could you please explain what is this, how do I detect it, and how to fix it. 

I know that this topic is related to GCP, by the way, I didn’t use any GCP in my flights, I am working with direct georeferencing of a Phantom 3 Pro.

Thanks.

Hi Edison,

Sure! Could you upload your Quality Report to our OneDrive here? We’ll analyse your results, detect problems and give you the best practices for processing your datasets. 

Best 

1 Like

I uploaded the Quality Report,

Thank you very much.

Hi Edison,

Thank you for sharing your Quality Report with us. To learn more about that effect, please have a look at our explanation in the Pix4Dmapper troubleshooter

This error can indeed be eliminated by marking GCPs in the project. However, this additional source of information isn’t always available. In that case, changing the processing options would be a workaround. In your dataset, the “bowl effect” doesn’t exist or is not visible much (maybe in the rayCloud is more noticeable). Nevertheless, the overlap between images isn’t sufficient; hence, the software couldn’t have computed many matches. Therefore there’re some issues we can improve by applying another processing options which at the same time can fix the “bowl effect” if present. I advise you to process your dataset one more time and include the following changes:

  • Initial Processing/General: Image scale: 1/2 (Half image size)
  • Initial Processing/Matching: Use Geometrically Verified Matching
  • Initial Processing/Calibration: Internal Parameters Optimization: All Prior

Once you have tried this suggestion, then please respond with an update so we can see if the results are accurate ;-) 

Best!

Thank you very much for your help. I uploaded to the link the second report. Actually, this is the important one, I made a flight at 66 meters of altitude but I got some uncalibrated cameras. I tried all the initial processing options in order to improve the keypoints matching but I still got some uncalibrated cameras (2%). I included some images taken at 120 meters of altitude and it helped a little bit (however, it reduced the GSD in 0.02 cm, not a big problem). 

Could you please check this report and let me know about the ‘bowl effect’, and how you recognize it. Remember I did not use GCPs. 

Thanks for all your help,

Edison

Hi Edison,

Did you try to reprocess your dataset with larger image scale? Half image size 1/2, for example? Remind the rest of the options in Step 1. 

Because of the fact the area, you are trying to reconstruct is covered by forest with lots of vegetations the software has problems finding matches between images. The larger scale should help overcome this issue. That is also the reason why 6 separate blocks are detected. The dataset is smashed. Therefore, the reconstructions won’t be all-encompassing with the highest quality. 

I would also advise you to take a subset of your data where most of the problems occur and reprocess them separately. I believe you can find the right processing options for those images.  

Let me know how it goes.

Best!