遥感图像的显著-概要特征提取与目标检测

Extraction of saliency-gist features and target detection for remote sensing images

针对巨幅遥感图像的目标检测问题,提出了一种基于显著-概要特征的遥感图像自动目标检测算法.采用滑动窗口将巨幅遥感图像划分为若干个小尺度的区域,针对各个小尺度分块图像,借鉴人类视觉生理功能特性之原理,提取其显著特征和概要特征,其中的显著特征代表了图像中的显著信息及显著区域空间分布和关联信息,概要特征可从整体上反映该区域的背景/目标关联信息.通过对分块区域图像的分类鉴别以实现目标检测.实验结果表明:此方法能以高可靠性和高精确度检测出巨幅遥感图像中的目标.

An automatic approach to detect and classify targets in high-resolution broad-area remote sensing images is explored,which relies on detecting statistical signatures of targets,in terms of a set of biologically-inspired low-level visual features.The broad-area remote sensing images were first cut into small image chips with slide window,which were analyzed in two complementary ways： attention/saliency analysis exploits local features and their interactions across space,while gist analysis focuses on global non-spatial features and their statistics.Both saliency and gist feature sets were used to classify each chip as containing target or not,through using a support vector machine.The proposed algorithm outperformed the state-of-the-art HMAX algorithm in the experiments and thus can be used to reliably and effectively detect highly variable target objects in large scale remote sensing image datasets.