Calibrating nightlight images (dark low texture background) with optional precise IMU positioning data

Hi,
we have done an airborne flight taking images (approx. 1500 images) at night to analyze the nightlight in a city. Example Image below.

The images are taken with a commercial NikonD3 camera (nikon raw images .nef) mounted on a stabilized platform, and we have simultaneously acquired differential GPS and inertial data.

Until now I have played around with several settings in the first calibration step, but the images could not be calibrated. (I have loaded the raw files with the initial GPS coordinates of the Nikon camera.)

Are there any options I should try?

We also have accurate dpgs locations and angles in an external file. Would it help to write the locations and angles to the exif files of the images? Or can I include external location and angle information somehow in the process?

Any help, is highly appreciated. Thanks.

You don’t have to write the exterior orientation values to the exif data. Pix4DMapper can import a csv with the x,y,z, omega, phi, kappa values:
Image geolocation file
Then I would try using the accurate geolocation and orientation calibration method in step one.

1 Like

Thanks. I will try that and will report here.

Hi @magnus, as @Andrew_Milanes mentioned, you can import the image geolocation and orientation using the From File… option in Pix4Dmapper.

Keep in mind that:

Let us know how it goes,

Hi,
I tried your advice but it is still not able to calibrate the images.

I used geolocation file (with: imagename,latitude [decimal degrees],longitude [decimal degrees],altitude [meter],omega [degrees], phi [degrees], kappa [degrees])

Then I used the accurate geolocation and orientation calibration option.

I get the following error:
[Error]: [BundleAdjuster][covariance] no projections errors ! aborting.
[Error]: Error e0046: Processing failed. No calibrated cameras.

What else should I try? I have the feeling that the nighttime images have too low “image quality” in terms of content/texture. The image shown at the top is one of the images with more content, there are images with much less light.

Any help is much appreciated. Thanks.

@magnus, indeed, it could be that the image content does not allow finding enough common points and that is why the calibration fails.

Most common issues for non-calibrated images are:

  • The image quality is not sufficient (blurry images, no texture, etc.).
  • There is no overlap between the images.

You can give it a try by changing the following processing options, but I, unfortunately, cannot guarantee that it will work due to the content of images:

  • Reduce the Keypoints Image Scale to 1/2 (Half image size) or less
  • And/or select 10’000 as Number of Keypoints in Process > Processing Options > 1. Initial Processing (select the Advanced checkbox)> Calibration > Targeted Number of Keypoints.

Best,

@Blaz, Thanks again for the reply.
I have tried the proposed parametrization with different input image sets, but it is still not working.
thanks anyway.