Utilizing Multiple Point Cloud Scenes for Precise Robotic Bin-Picking Tasks

This study proposes a novel scan-matching approach for accurate object pose estimation in robotic bin-picking using only a single low-cost RGB-D camera. Instead of relying on multiple sensors, the method reconstructs a richer point cloud by combining views from different camera angles, improving visibility in cluttered or occluded environments. The Iterative Closest Point (ICP) algorithm is then applied to align object models with the reconstructed scene to recover precise poses. Experiments show that the proposed approach reduces translation and rotation errors by around 70%, raises pose estimation accuracy to 97.5%, and enables an average bin-picking success rate of 97%, even for complex object geometries.