DeepVCP: An End-to-End Deep Neural Network for Point Cloud Registration
DeepVCP(Virtual Corresponding Points) is an end-to-end learning-based 3D point cloud registration framework that achieves comparable registration accuracy to prior state-of-the-art geometric methods of aligning two different point clouds.
Instead of implementing other keypoint based methods where a RANSAC procedure is usually needed, the implementation of various deep neural network structures is done to establish an end-to-end trainable network. The keypoint detector is trained through this end-to-end structure which enables the system to avoid the inference of dynamic objects and leverages the help of sufficiently salient features on stationary objects, thereby achieving high robustness.
Main Contributions
-
First end-to-end learning based point cloud registration framework.
-
Generation of corresponding points are done based on learned matching probabilities among a group of candidates which improves registration accuracy.
-
KITTI & Apollo-SouthBay datasets are used to validate it’s efficiency.
Results demonstrate that it achieves comparable registration accuracy and runtime efficiency compared to state-of-the-art geometry-based methods, but with higher robustness to inaccurate initial poses.
Low registration error and high robustness of this method makes it suitable for substantial applications based on point cloud registration.