Motion Data Preprocessing in Robotic Applications

Abstract

There are numerous approaches to processing images of moving bodies.  As to robotic applications call for real time algorithms the redundant data obtained from the input image should be filtered out and implemented with deep knowledge of hardware specifications.  The paper demonstrates a new approach to real-time motion classification. Partial problems resolved in this paper cover the fast motion detection, edge detection, corner detection and finally motion classification.  The motion, edge and corner detectors are optimized with respect to effective use of the memory and excluding redundant data.