Fast and Accurate 6-D Object Pose Refinement via Implicit Surface Optimization

Abstract

Aligning a point cloud to a fixed 3-D model is a crucial task in many applications, such as 6-D pose estimation for robotic grasping. Typically, an initial pose is estimated by analyzing both the point cloud and the 3-D model, after which the iterative closest point (ICP) algorithm is used to refine the pose, reducing large errors and improving accuracy. In this article, we propose an accurate and efficient alternative to the ICP. Our method encodes the fixed 3-D model into an implicit neural network, which is trained offline as a one-time process in just a few minutes, requiring only the CAD model of the object. The network takes the point cloud and pose as inputs and outputs the signed distance field (SDF) value. By minimizing the absolute SDF value with the fixed point cloud and network weights, while optimizing the pose, we obtain the final precise alignment. The key advantage of our method is that it eliminates the need to explicitly establish one-to-one correspondences between the point cloud and the 3-D model, a necessary step in the ICP and its variants. This enables our framework to avoid local optima and makes it more robust to challenging conditions such as large initial pose gaps, noisy data, variations in scale, occlusions, and reflections. Furthermore, the end-to-end network of our framework offers significant runtime efficiency. We validate the superior performance of our approach through extensive comparisons with various ICP variants on both synthetic and real-world datasets.