Abstract
PointNet has recently emerged as a popular representation for unstructured point cloud data, allowing application of deep learning to tasks such as object detection, segmentation and shape completion. However, recent works in literature have shown the sensitivity of the PointNet representation to pose misalignment. This paper presents a novel framework that uses the PointNet representation to align point clouds and perform registration for applications such as tracking, 3D reconstruction and pose estimation. We develop a framework that compares PointNet features of template and source point clouds to find the transformation that aligns them accurately. Depending on the prior information about the shape of the object formed by the point clouds, our framework can produce approaches that are shape specific or general to unseen shapes. The shape specific approach uses a Siamese architecture with fully connected (FC) layers and is robust to noise and initial misalignment in data. We perform extensive simulation and real-world experiments to validate the efficacy of our approach and compare the performance with state-of-art approaches.
Code & Video
Results
Qualitative Results
Registration results of various object categories from ModelNet40 dataset with noisy source point clouds. The above results show that the results of our iterative-PCRNet (Red colored points clouds) are aligned much better than the iterative closest point (ICP) method (Green colored point clouds) which is a state of art conventional registration method.
Registration of chair point cloud taken from Stanford S3DIS (Link) indoor dataset. CAD model shows the template data fromModelNet40, purple points is from S3DIS dataset, red points represent iterative PCRNet estimates, while the green ones represent ICP (iterative closest point) estimates.
|
|
|
|
Qualitative results of chair point clouds from S3DIS indoor dataset by Stanford. We registered the chair point clouds inside the rooms using our iterative-PCRNet method and then these chairs are replaced using various available CAD models. It clearly indicates that PCRNet shows interesting applications in the furniture industries and AR/VR technology.
|
Quantitative Results
Statistical comparisons of our PCRNet method with ICP (conventional method) and PointNetLK (learning based method)
|