A Hybrid Approach toward Scrater Detection Based on Illumination Direction, Texture Analysis and Interpolation

摘要

Impact craters are one of the most important topographical features of extra-terrestrial surfaces. While they serve as ideal landmarks for navigation as well as landing, they do pose problems by also turning out to be hazardous areas for spacecrafts desiring to land. Unmanned landing requires accurate estimation of the surface conditions so that unsafe regions may be identified and avoided. A standard deviation filtering and texture analysis-based scheme proposed recently, has been observed to have a low detection rate for small craters. This article uses edge-preserved interpolation to improve on the above scheme combining it with a novel direction of illumination based approach resulting in an efficient hybrid algorithm. The approach yields an average 90% true positive rate for tests run on real images of lunar and Martian surfaces. The major contribution of this algorithm lies in its over 94 % true positiverate for craters, in nadir-viewing images.