Depth map with Raspberry Pi

Question

Is it possible to get two images from the Raspberry Pi camera mounted on a remote controlled bot and have them sent to a computer through Wi-Fi and process the images in the computer to generate a depth map?

All this is to be done in a very short time so that the robot can be helped with its locomotion without making it completely autonomous.

Answer 1

Yes, it is possible. This is a form of stereo vision. You will need an accurate model of how the robot moved between frames. Then you can use stereo vision techniques to calculate the disparity.

Answer 2

I am currently using the Compute Module IO board for my undergraduate project and up to now, the best I can come up with is a refresh rate of 2Hz implemented using c++.

Mr. David Barker at Argon design was able to implement it at 1Hz using C and 12 frames per second by using machine language on the video core.

Mr. Nalan Karunanayake was able to implement stereo vision using a model B+ and two usb webcams set at 10 frames per second in c++.

I read an IEEE paper if I am remembering it correctly being able to sync both camera modules using v4l2 drivers (Raspberry PI Based Stereo Vision For Small Size ASVs^†).

^{† Requires IEEE Xplore subscription or purchase.}