Real-world objects have depth, width and height, and automated systems such as robots need to “see” in these three dimensions if they are to operate successfully. 

Machine vision systems composed of a camera attached to a computer running special image processing software give robots the “sight” they require. Ever since machine vision gained mainstream attention in the 1980s, however, one of the biggest challenges facing machine vision system designers has been how to best reduce the amount of data that needs to be processed to correctly locate, inspect and analyze an object.