How to generate 3d image from a single 2d image using backprojection of pixels
I have a 2d binary image of segmented lanes i want to create a point cloud of this image. I do not understand the math of using a pinhole camera model to pixel back projection in 3d. Can you please help me convert this image to point cloud where im assuming a flat plane and projecting the pixels on it. On many sources i noticed everyone is suggesting me to this method so please help me with it.
Asked by Sotte17 on 2019-03-14 11:43:28 UTC
Comments
Can you include what you've tried and the resources you've looked into thus far?
Asked by stevemacenski on 2019-03-14 12:31:13 UTC
I haven't tried much till now but I have seen a lot of resources like how they are implementing in this GitHub code https://github.com/RoboJackets/igvc-software/blob/master/igvc_perception/src/cnn/cnn.py
Asked by Sotte17 on 2019-03-14 17:04:28 UTC
I'd recommend understanding the pinhole model of a camera before diving into anything that complex. What resources have you looked at for pinhole cameras that confused you?
Asked by stevemacenski on 2019-03-14 17:59:04 UTC
Well I saw a video on YouTube but It wasn't clear to me. Then I saw some other documents like then opencv documentation for 3d reconstruction
Asked by Sotte17 on 2019-03-15 01:37:12 UTC
Please try to look more into it, I think you'll find your question quickly answered by doing a little research. Here's an example of a pretty good discussion on the model: https://web.stanford.edu/class/cs231a/course_notes/01-camera-models.pdf
Asked by stevemacenski on 2019-03-15 01:40:25 UTC