Line-based registration for UAV remote sensing imagery of wide-spanning river basin

Abstract

With the development of Unmanned Aerial Vehicles (UAVs) remote sensing technique, some researchers began to apply the UAV imagery to investigate Yellow River's dike hazards. There has been an urgent need for establishing automatic and accurate registration techniques of UAV remote sensing imagery. High-precision image registration will enhance the reliability of the investigation. As the Yellow River is very wide, it is very difficult to find enough Ground Control Points (GCP) and the limited GCPs are usually distributed unevenly. Therefore, the traditional point-based image registration approach cannot satisfy the requirements of imagery registration. However, the river basin contains linear features richly. This paper proposes a line-based registration approach for UAV remote sensing imagery of the wide-spanning river basin. This method regards the segment straight line as the registration primitives. After detect the primitives of input images, we estimate the parameters of transformation model between reference-image and registering-image with Modified Iterated Hough Transform (MIHT). Experiments using the UAV imagery of Ning-Meng section of Yellow River conclude that this is a robust method for registration for UAV remote sensing imagery of wide-spanning river basin.