Multi-sensor Image Registration for Precisely Locating Time-sensitive Objects

摘要

Finding time-sensitive objects in time and precisely locating them in real time is the basis of successful operations. This paper analyzes the characteristics of images from space-based platforms such as satellites, and images from aerial platforms such as UAVs, and then presents a method for precisely locating time-sensitive targets. The idea behind this method is that, because the satellite images from intelligence image database have accurate geodetic coordinates associated with every pixel, aerial images can obtain the geographical coordinates of high accuracy by registration with the satellite images. The key of this method consists in alignment of the two kinds of images. To do so, this paper presents a two-stage image registration algorithm. The first stage performs image stitching for aerial images so that the resulting images can include more structural information of the scene. In the second stage, the resulting images are aligned with the geo-referenced images from the image intelligence database. Experimental results show that the proposed algorithm is effective and robust in case of scene variations.