Support Website Contact Support Blog

low number of calibrated images

Using the eBee classic drone and Sony WX RGB camera, I obtained aerial images of forested area. But after initial processing quality is not sufficient and number of calibrated images are low. I know that when working on dense vegetation, on initial processing stage indicating image scale less than 1 improves quality of results. However, on 0.5, 0.25 and even on 0.125 scale I could not get desirable quality (respectively only 29%, 27% and 24% of the images were calibrated).
I suppose mixed landscape would be better, but the subject of the mission was that dense forest.
Is there anything I can change in processing settings that could yield better results?

I am attaching report (with the image scale 0.25 - others also available), and also a sample image of the forest.
Thank you.

0.25_image-scale.pdf (2.5 MB) DSC02456_512

Does you eBee have PPK? If try changing the calibration method from Standard to Accurate Geolocation and Orientation.

1 Like

No, it does not. It’s the eBee Classic. But pics are geotagged (though without high accuracy) when processed “postflight” using the soft by the drone manufacturer. Still, in the pix4Dmapper, most of them are without any geotag.
To my knowledge, “accurate geolocation” option in processing is for pics obtained by a drone with RTK/PPK.

Indeed, the “accurate geolocation” option is for images with very good position and orientation accuracy.

Adjusting the image scale would be my first suggestion. I am sorry to hear it did not work for you.
Have you seen this article: How to improve the outputs of dense vegetation areas?

Flying higher and increasing the overlap are other possible ideas. Unfortunately, nothing else comes to my mind that is not described in the above mentioned article.
I hope it helps :slight_smile:

Thank you Christina. Next time I will fly higher to get better results. This time nothing helps and what I have is maximum only 29% of images get calibrated (although all the pics have been geotagged, most of them are not taken by pix4D ).