Pose Estimation for 3D Work Piece Using Differential Evolution Algorithm

In industrial fields, precise pose of a 3D object is the prerequisite of the subsequent tasks like grasping and assembly, thus many researches on accurate pose estimation of a 3D object are explored over the last decades. To get the pose of a 3D work piece from the 2D image data is a challenging task in industrial applications. This paper proposes a fully automated pose estimation system which is capable to estimate the accurate model and pose of a 3D work piece that can well match the 2D image data. This is achieved by representing the above problem as an optimization problem aiming at finding the accurate model parameters and pose parameters of work piece by minimize the difference between the real 2D image and the hypothetical 2D image that produced through the given parameters from 3D image. Due to the coupling of the unknown model and pose parameters and the discontinuity of the objective function, the above optimization problem cannot be solved through traditional optimization approaches. Hence, we utilize a heuristic optimization strategy - Differential Evolution to cope with the problem. The experimental results demonstrate the effectiveness of the proposed method.