Determination of occlusion areas in high resolution remote sensing data

Abstract

In this paper we discuss different methods to determine occlusion areas in high resolution images. Occlusion in general occurs if the local slope angle of the terrain surface is larger than the off-nadir look angle of the sensor. Here, the sensor line of sight intersects the digital surface model more than once, leading to duplication artefacts when rectifying the image data. This especially happens for high buildings or similar man made objects, trees or steep terrain. In principle there are two approaches to evaluate the relationship between a target point on ground and the respective image pixel, which are based on a backward geocoding and a forward geocoding algorithm, respectively. Either of these approaches first has to detect and determine the multiple intersections of the line of sight and then has to identify which of them is addressed to an occluded area. The algorithmic characteristics as well as the performance of both approaches with regard to computational effort and quality of results are discussed in this paper.