Pose estimation of top object on stack of elliptical objects based on ellipse fitting and Kalman estimator

Abstract

Abstract:- This paper presents a top object selection and its pose determination algorithm which can be used to pick the top one among a stack of overlapped elliptical objects provided on a moving conveyer belt. For this purpose, an ellipse fitting algorithm is developed to find the ellipse shapes and their parameters in the image. It also determines the top object among them and its pose by analyzing the edge patterns. Based on the object’s speed and pose measured in the image frame, the Kalman estimator determines the pose of the object at the moment of pick-up with considering the noises caused by the motions of robot and conveyer belt. Once the object’s pose to pick is determined in the image frame, then its corresponding robot’s pose in the world frame is determined using the image Jacobian. The proposed algorithm implemented and installed in the 6DOF Denso robot has shown a prominent performance in the experiments.