RGB-D Camera-Based Object Grounding Surface Estimation System

3D point cloud data has attracted growing attention in various fields. The identification of grounding surfaces from point cloud contributes for numerous practical applications including interactive systems and infection prevention. However, it is difficult to distinguish a target object from other objects with similar shapes. RGB cameras can recognize targets with similar shapes, but cannot accurately gauge the contour of the grounding surface. There have been no studies on seamless methods combining RGB and depth data to estimate the grounding surface of a target object. To address this problem, this paper proposes a grounding surface estimation method using RGB-D data. It first identifies the target object from RGB data. Then, the grounding surface of the target object is estimated from point cloud data. The grounding contour of the target object is accurately estimated even when several objects with similar shapes are on the physical surface. The performance of the proposed scheme was evaluated from experiments using Intel RealSense D455. We confirmed that the proposed scheme achieves identifying the grounding surface, even in the presence of non-target objects.