NOVEL ROBUST APPROACH FOR COARSE DEM AUTOMATIC MATCHING

This study proposes a novel method for improving the performance of digital elevation model (DEM) matching and for solving the initialization problem of the iterative closest point (ICP) algorithm. The principles of computer vision were applied to extract, label, and match DEM feature points for coarse DEM matching. The DEM feature points were further used to initialize the ICP registration algorithm to estimate the transformation relationship between two DEMs. Then, learning optimized local terrain binary (LLTB), a new binary descriptor for matching DEM feature points, was developed to improve the efficiency of coarse DEM matching. Experimental results indicated that LLTB is a highly efficient, discriminative, and stable descriptor of local DEM information. This study contributes to autonomous aircraft positioning and navigation research by matching the DEM data of flight routes with the baseline DEM. Composite navigation methods can be developed by combining the proposed method with inertial and satellite navigation methods that correct the drifts of inertial navigation systems. Moreover, the proposed method can remedy the limitations encountered in satellite navigation when GPS is unavailable.

Keywords: computer vision; coarse DEM matching; LLTB; ICP; feature descriptor