I’m experimenting with making a 3D model of a house using the insta360 Pro 2 camera. The camera has six 200 degree fisheye lenses, which is of course difficult for RC to handle. I have not stitched them together into one 360 photo, so each setup has 6 different photos. I’m wondering if anyone could help me figure out what parameters I should use to maybe be able to get a model out of them.
Insta360 didn’t want to give me the lens distortion parameters but they could tell me this:
“The actual focal length of Pro2 is 1.88mm, with a cropping factor of 5.619.
Since the lens is fisheye, the image produced by the actual focal length of 1.88mm is cropped and enlarged, equivalent to an image taken with a lens of a focal length of 1.88mm multiplied by 5.619, which is 10.57mm. Therefore, the focal length is 10.57mm.“
Hi @amfram,
what kind of results are you getting without any changes in the settings?
For such cameras is a good way to use the Division camera model (it would be good to know the Radial 1 parameter).
Regarding to the focal length, is it in 35 mm format? If so, you can add that to the Prior calibration section and use that for the alignment.
I don’t know if it’s based in a 35mm format, nor the radial 1. It’s difficult to get detailed information from insta360. Here is the result from only using the Division model. As you can see it only produces a lot of small components with a lot of distortion. The sparse point cloud looks like a sphere.
I have also scanned the whole house with a Leica RTC360 LiDAR scanner, can that be of help even if i don’t know the parameters of the Pro 2 camera? I wan’t to make a photorealistic model, and only using the LiDAR scanner won’t get me that. Is it better to crop and try to flatten out the distortion in a different software before processing it in Reality Capture?
What is the overlap between your images?
Can you show, how the image looks?
As you can see, this can be problematic. You can also try to improve the alignment settings a little bit (add more features, change the detector sensitivity, …). But still it can be problematic as on the lens front, there’s a lot or recommendations to stick with a focal range of around 35mm (from 24mm to 50mm)
The scan can help, but there needs to be a good overlap between the data. Or you can use the control points.
