Bojan, You could contact Shrinivas Pundlik and Gang Luo at the Schepens Eye Research Institute in Boston, MA, USA. They have a single-camera, obstacle-detection (warning) system for people with vision impairments.

Thanks, Russell. Seams that for indoor application and short distance up to 10 m. My application is in really dynamic outdoor environment and distances up to 50m. Also they are not using any machine learning. Doubt thats applicable for my scenario and quite robust enough. Any other suggestions?

Bojan, Do not be restricted by what you can currently read, as published research is going to be at least a few years behind current work. To my knowledge they are using machine learning in one approach (I attended a talk yesterday). Just because they have reported about helping people with vision impairments (for which a 10m distance is more than is needed), I do not see a reason that their technique would not work at longer distances. There is no reason that it would not work for space travel. They have looked at it for automobile use (examples of roadway images and distant natural scenes were used in the talk). The restricting factor in any application is the ability to detect objects or aspects of objects. Thus, if the object is hard to see against its background (e.g. contrast, clutter), then a single-camera approach will have difficulty. As I wrote earlier, I recommend contacting them directly.

thanks. I will