Correlation based object-specific attentional mechanism for target localization in high resolution satellite images

摘要

Attentional Mechanism or Focus of Attention is the front end of object recognition systems with the task of rapidly reducing the search area in the image. In this paper we present correlation based template matching as an attentional mechanism for high resolution satellite images. We experimentally show that despite intra-class variations and object transformations, correlation based template matching can be deployed as attentional mechanism. Different image variants like gradient magnitude and gradient orientation are also compared for correlation matching. Based on the experiments a threshold selection mechanism is given.