Fast image capture and vision processing for robotic applications

Abstract

This chapter details a technique to significantly increase the speed of image processing for robot identification in a global-vision based system, targeted at real-time applications. Of major significance are the proposed discrete and small look-up tables for Y, U and V color thresholds. A new YUV color space has been proposed which significantly improves the speed of color classification. The look-up tables can be easily updated in real-time and are thus suitable for adaptive thresholding. The experimental results confirm that the proposed algorithm greatly improves the performance of the image processing system. The results are compared with other commonly used methods such as a composite look-up table which is indexed using RGB pixel values. © 2008 Springer-Verlag Berlin Heidelberg.