SphereNet: Learning a Noise-Robust and General Descriptor for Point Cloud Registration

Point cloud registration aims to estimate a transformation that aligns point clouds collected from different perspectives. In learning-based point cloud registration, a robust descriptor is crucial for achieving high-accuracy registration. However, most existing methods are susceptible to noise and demonstrate poor generalization ability when applied to unseen datasets. Motivated by this, we introduce SphereNet to learn a noise-robust and unseen-general descriptor for point cloud registration. In our method, first, the spheroid generator builds a geometric domain based on spherical voxelization (SV) to encode geometric information. Then, the spherical interpolation of the sphere is introduced to realize robustness against noise. Finally, a new spherical convolutional neural network (CNN) with spherical integrity padding completes the extraction of descriptors, which reduces the loss of features and fully captures the geometric features. To evaluate our methods, a new benchmark 3DMatch-noise with strong noise is introduced. Extensive experiments are carried out on both indoor and outdoor datasets. Our results demonstrate that SphereNet achieves an increase in feature-matching recall of more than 25 percentage points (pp) on 3DMatch-noise under high-intensity noise. Moreover, SphereNet establishes a new state-of-the-art performance on the 3DMatch and 3DLoMatch benchmarks, achieving 93.5% and 75.6% registration recall (RR), respectively. Furthermore, SphereNet exhibits superior generalization ability on unseen datasets.