A Holistic Vision-based Hazard Detection Framework for Asteroid Landings

Abstract

A holistic vision-based hazard detection framework for asteroid landing is introduced in this paper. The proposed holistic conjecture takes advantage of the Hough Transform method and decision forests, shares the simplicity and wide applicability of the Hough transform but bypasses the problem of obstacle multi-size detection and permits detection of multiple objects. The framework is conducive to hazard avoidance, satisfying the safe bound of the lander and the exploration task, and adaptive to environment, autonomous without additive complex filtering. The results show that the framework provides detection and localization performance close to that of human calibrations and can be further improved when larger and more diverse datasets are available.