Sequential Filling Method for Internal Holes in Road Elevation Map Based on Bilateral Filtering

Abstract

Road elevation maps have practical applications in fields such as civil infrastructure and autonomous driving systems. Higher-quality elevation maps are demanded in these fields. However, holes may occur within the elevation maps and result in incorrect applications. It is important to research the filling methods for internal holes in road elevation maps. This paper proposes a hole-filling method for road elevation maps based on bilateral filtering. First, the information quality function of the invalid grid neighborhood for the filling order of the bilateral filtering is constructed; second, the parameters of the function are optimized by the active gradient projection method; then, an adaptive repair window is made according to the hole-filling path formulated by the information quality function; finally, the internal holes of the elevation map are filled sequentially based on the total information quality of the grid to be repaired and the adaptive repair window. Experiments are carried out by the Tum data set and the actual outdoor data respectively. The experiment results show good accuracy, robustness, and real-time performance of the proposed method.