GLOBAL IMAGE MATCHING BASED ON FEATURE POINT CONSTRAINED MARKOV RANDOM FIELD MODEL FOR PLANETARY MAPPING

摘要

In planetary exploration missions, high accuracy mapping from orbital and rover images is fundamentally important for scientific investigation, landing-site selection, precision landing, and rover navigation. Stereo image matching is a critical technique for 3D planetary mapping. A new global image matching method is presented based on feature points and Markov Random Field (MRF) model. The method extracts feature points, predicts disparity range, minimizes energy function of MRF, and consequently gets dense matching results. Experimental results using rover images from the Mars Exploration Rover mission and orbital images from Chang'E-1 lunar mission are presented to demonstrate the effectiveness of the proposed method.