ACCURACY OF LOCALIZATION OF AREA AND LINEAR OBJECTS IN SPACE IMAGES
V. A. Grishin
Space Research Institute, Russian Academy of Sciences, Moscow, Russia
Keywords: space image registration, accuracy of matching algorithms, area and linear landmarks, vector reference images, Cramer-Rao estimation
Abstract
Image registration methods are widely used in processing satellite images of the Earth. This solves the problem of establishing the correspondence of images of the same landmark obtained from different distances, different angles, under different lighting conditions, at different times of the year, etc. This article presents a method for assessing the potential accuracy of localization of area and linear landmarks of significant sizes, which can be considered flat given the altitude of the orbit from which the survey is made. The accuracy of localization is determined depending on the shape of the vector reference image, its size, accuracy, and the accuracy of localization of the boundaries of the observed image of the object. It is shown that, under favorable conditions, it is possible to obtain subpixel accuracy of localization of area and linear objects of significant sizes.
|
