Vision based robot navigation: Naive Implementation

[timeshiftersvk]

Can you please tell me, how this works? It seems like you only use 1 image from first camera for more images from second camera. Then you match SURFs and compute relative position of (old) image from first camera to the position of second camera (robot).

Do you use SURF pairs in computation of disparity map? If so, can you please tell me, which algorithm do you use?

[scyscyao]

@timeshiftersvk I just implement the find_obj and stereo samples in OpenCV and integrated them together. The disparity map is created by BM algorithm

[timeshiftersvk]

@scyscyao Hmm, it looks way better than on my camera system (2 cheap webcams :D ).

[scyscyao]

@timeshiftersvk haha, because I use bumblebee2 which is specially designed for stereo vision. For webcams, you need to fix them and do rectification every time

[timeshiftersvk]

@scyscyao ...and also, thank you for your reply.