A Segment-Wise Gaussian Process-Based Ground Segmentation With Local Smoothness Estimation

In terrestrial and extraterrestrial environments, the precise and informative model of the ground and the surface ahead is crucial for navigation and obstacle avoidance. The ground surface is not always flat, the functional relationship of the surface-related features may vary in different areas of the ground, and the measurement of point clouds derived from the ground surface does not necessarily reflect smoothness. Thus, ground-related features of the ground surface must be calculated based on local or point estimates. This paper proposes a segment-wise Gaussian Process (GP)-based ground estimation method with input-dependent smoothness estimation. The value of the length scale is estimated locally for each data point resulting in a more precise estimate for the rough scenes while being not computationally complex and robust to under-segmentation, sparsity, and under-represent-ability issues. The segment-wise task estimates a partial continuous model of the ground for each radial range segment. The simulation results show that it can give a precise continuous estimate of the ground surface in rough and bumpy scenes while being fast enough for real-time applications.