Real Time and Robust 6D Pose Estimation of RGBD Data for Robotic Bin Picking

In this paper, we propose a real time and robust method to address the issues of 6D pose estimation for robot bin picking utilizing a low cost 3D sensor. RGB-D data have been widely in industrial applications to deal with 6D pose estimation problems. Different from previous methods of registration-based with time-consuming and point cloud information only with poor accuracy, our approach uses a pinhole camera model and the geometric relationship to correlate the point cloud data and RGB pixels, which is faster and almost same accurate compared with registration based methods. Experimental results show that our method is computationally efficient and can reach an approximate centimeter accuracy. Furthermore, it is also much lower cost and easier to implement when Compared with anterior methods.