Real-time surface slope estimation by homography alignment for spacecraft safe landing

This paper offers a real-time approach for simultaneously determining spacecraft motion and multiple 3D planar surfaces for spacecraft safe landing. The approach contains three algorithms: a multiple homography alignment algorithm, which constructs homographies under unified epipolar geometry; a closed-form motion estimation algorithm; and a simple routine for surface slope estimation. This approach has three significant advantages: first, it works well for both a simple planar scene and more complex 3D world containing many smaller planar surfaces; second, it decomposes a very large optimization problem into much smaller sub-problems that are computationally less expensive. This computational advantage means that this approach can be exploited in real time robotic operations such as during time critical spacecraft landing. Finally, this approach is very reliable and accurate. The effectiveness of this approach is determined quantitatively through extensive simulations and qualitatively with actual images.