The Image Quality Is Affected By The Three-dimensional Positioning Accuracy

EOS (Earth Observation System) is observing the earth by remote sensors which are carried by spaceflight, it is very significant. For example, it can be used in military reconnaissance and resources exploration; it also can be used to get stereo pairs of images, which are used to make maps and get the DEM of regions, and so on. These products are all based on processing and computation of remote images. So, good quality of images can fulfill the requirement of more high-precision mapping products. This paper is written in view of follow questions:how to evaluate quality of remote images, what are the parameters that determine quality of remote images, and how these factors affect precision of mapping products.This paper makes a research about the relationship between factors (MTF and SNR, which are related to image quality) and precision of stereo positioning. This paper introduces key technologies about stereo positioning at the beginning; because the result of image registration is closely related to precision of stereo positioning, we select appropriate image registration algorithm. First, get corners information using Harris corner detection method, then get rid of error matched pixels, at last, locate homologous image points to sub-pixel using least square algorithm; in the paper, it introduces two methods which are used to compute MTF of on-orbit remote sensing images and make a comparison between them, In order to compute precision of stereo positioning, this paper presents the concept of Height Function, which is used to get the real 3D model coordinates, because we can not know it after getting the 3D coordinates of model points by using the technology of stereo positioning.In the end of this paper, it shows the relationship between MTF (SNR) and precision of stereo positioning and proposes the technique index of remote camera, which supports the research of satellite system. There are two kinds of models in this experiment, the result of which can be used to confirm each other. The results of these two models are proved consistent.