Ground Target Localization Algorithm Based on UAV Aerial Image Analysis

Abstract

Ground target localization is of great importance in UAV aerial image analysis and computer vision, and many researchers have been studied on this topic. In this paper, we focus on the problem of ground target localization using the Multi-feature fusion technology. In this work, visual features are extracted from UAV images, and then visual similarity is calculated, on the other hand. Afterwards, the coordinate in coarse positioning result is estimated by the radiation transformation model and multi-feature fusion. The main innovation of this paper lies in that we effectively integrate different types of image visual features, such as color moments, color histogram, color correlogram, GLOM texture features, and Tamura texture features. Experimental results demonstrate that effectively integrating all these features, the accurate Ground target localization results can be achieved.