Vision system adapted to a mobile tracked robot

This research is focused on the implementation of a vision system intended for a mobile tracked robot, based on three different detectors and feature descriptors which have been analyzed to determine which gives the best performance. The metrics used for comparison purposes are: time processing, resistance to noise and invariance to rotation. Experimental tests conducted demonstrated that the Oriented Fast and Rotated Brief (ORB) approach is the fastest, providing similar results in terms of noise resistance and invariance to rotation showing good performance in a real world environment.