Structured-Light-Based 3D Scanning System for Industrial Manipulator in Bin Picking Application

Abstract

This paper developed an industrial manipulator integrated with a structured-light-based 3D scanning system using Gray code and phase-shifting patterns, implementing the bin-picking problem. The procedure of the vision-based robotic system developed in this paper is described as follows: First, the object is reconstructed in point cloud format by projecting a series of Gray code patterns followed by four-phase shift patterns. Point clouds of the object from different views are filtered and concatenated to obtain the full cloud. The full clouds, altogether with partial clouds, then are fed to PointNet neural network as a training dataset, whose estimated grasping pose can be used to pick up the objects from a bin. By using the estimated grasping pose, the robotic arm can reach the desired position and pick the object. Experimental results indicate that the developed vision-based system can reconstruct objects and grasp different-sized objects in the bin.