Hierarchical-Search to Reduce Search-Time of a Machine-Vision System

摘要

This paper presents a modified template matching technique that greatly reduces the time to find and identify the potential match candidates of an image from a frame by employing the hierarchical-search method. The search algorithm developed reports significant reduction in search-time over that for traditional template matching. Rotation and scale invariance have been added to provide the necessary robustness to the algorithm. Finally, relations between search-time versus size of templates, number of match-candidates and different hierarchical levels of search have been examined. The algorithm promises a potentially improved machine-vision over those using neural network/other soft computing tools in terms of speed, and those employing traditional template matching, in terms of robustness.