Support Website Contact Support Blog

low quality mosaic

Hi there,

I had a pastureland research project, using eBee classic equipped with SEQUOIA multi-spectral camera.
As a result of processing RGB images with Pix4Dmapper I got the orthomosaic which is not really high quality when I opened it in QGIS and zoomed in - please see attached screenshot to see what I mean.

I suppose the reason for that is somewhere in the processing (and not in image acquisition).

Could anyone advice how to set processing settings to get better quality mosaic?? Thank you.
I am attaching also screens showing the settings I applied while processing the RGB images of the project (which resulted in that mosaic).
Thank you.

step 1



step 2



step 3



Hi Giorgi,

For the first step can you try using custom image scale and set it to 1/2 instead of full image scale? We would also need the quality report. You can upload it to google drive and share it here.

Sequoia RGB sensor is a rolling shutter one, that can cause some issues. Make sure you fly high and also slow to keep the effects of rolling shutter lower.

Hi Momtanu,
Thank you for reply.
I will re-process as you have advised - with the scale set to 1/2.

here is the link to the report https://we.tl/t-xLovqEfz3M

Besides, I am including the screenshot also from the flight mission where overlaps and altitude are shown (just in case you need that as well).

Thank you.
regards, Giorgi

Hi Momtanu,
I reprocessed with the image scale 1/2, but not really better.
below please find the screenshot showing the same area re-processed and here is the link to the report https://we.tl/t-z61hPnAZwl

Thank you for your support.
Giorgi

Hi Giorgi,

How does the raw images look? Everything in the QR looks fine except the graph for keypoint matches. The software seems to have found very few matches of keypoints between the images, that might be the reason along with rolling shutter. Also, the RGB camera of sequoia is mainly designed as an add-on to generate overview images of agricultural fields. The RGB camera, therefore, cannot provide images in the same quality as dedicated RGB cameras can do.

On the data acquisition side, we suggest flying smoothly to reduce abrupt movements and vibrations because of the rolling shutter. The sensor is exposed and read line-by-line, instead of the entire sensor being exposed at once. This can lead to distortions when capturing fast-moving objects or when capturing the image while moving at a high speed. There are limitations to the software’s ability to account for the distortion that is introduced into images captured by a rolling shutter camera that is not effectively isolated from vibration.

You may be able to account for the uneven distortion in the RGB images by downscaling the Keypoint Image Scale to either 1/2 (which you already did) or 1/4 the original resolution.