Abstract: An algorithm for attention selection of active machine vision system was proposed.This algorithm employs the pyramidal representation to resolve the problem of calculation amount,it also employs the sequential templet and saliency map to combine the upper layer information, and integrates both the data driven and model driven mechanisms.We had done a lot of tracing experiments based on our attention algorithm which include monocular and binocular tracing for human being and physical objects under complex scene on the experiment platform.The experiment results prove the effectiveness and robust of this algorithm.

Key words: active machine vision, attention selection, pyramidal representation, sequential templet, saliency map