VLSI-implementable scheme for invariant object recognition

The authors describe the development of a VLSI smart sensor which borrows some mechanisms used by the human eye to generate abstractions of simple line drawings which are invariant with respect to scale and position. It is demonstrated by simulation that a corner-based abstraction method is suitable for generating signatures invariant with respect to scale and position. It is shown that such a sensor (1) is capable of real-time performance, (2) yields a considerably higher resolution than conventional schemes, and (3) is similar in some important respects to the human visual system. Two implementation schemes are described, and it is shown that they are capable of real-time performance. Limitations of the technique are mentioned, as is the simple extension of the method to gray-scale images