Image Based Autonomous Navigation for a Lander

摘要

The aim of the proposed work is to provide an autonomous navigation algorithm for a lander to a safe landing site with zero position errors. The algorithm uses the monocular images from the lander camera as an input and produces position and motion estimation. The algorithm used for image matching is Scale-Invariant Feature Transform (SIFT) algorithm which uses the moon craters for feature extraction and feature matching. Feature matching between consecutive images helps in obtaining SIFT flow of objects between images. The two images to match may contain object instances captured from different viewpoints, placed at different spatial locations, or imaged at different scales. The SIFT algorithm provides feature detection which is indifferent to scaling and to an extent to illumination as well. This proposed work takes cue from optical flow methods, which produces dense, pixel-to-pixel correspondences between two images. This paper proposes SIFT flow. It adopts the computational framework of optical flow, but by matching SIFT descriptors instead of raw pixels.