Integrating Edge and Shadow Constraints Into Dense Optical Flow for High‐Fidelity Asteroid 3D Reconstruction

ABSTRACT

Accurate and detailed terrain information is a prerequisite for the autonomous navigation and safe landing of asteroid probes. Inaccurate boundary feature description in dense matching algorithms has always been a challenge. To obtain a fine 3D terrain model with accurate boundaries, this study proposes a dense optical flow estimation algorithm for asteroid images that considers edges and shadows on the basis of the geological characteristics of the rubble‐pile asteroid. The method adopts a new irregular adaptive window generation algorithm with image edges and shadows as spatial constraints. Moreover, an adaptive window and dynamic weighting‐based optical flow tracking algorithm is established for epipolar images. A dense optical flow estimation algorithm framework that integrates the pyramid strategy, bidirectional optical flow, and the median filtering method is designed. Experimental results demonstrate that the proposed method yields dense matching results. Compared with traditional dense matching algorithms, the proposed method has remarkable advantages in detail performance and algorithm robustness.