ICP-4D: Bridging Iterative Closest Point and LiDAR Panoptic Segmentation
This work addresses computational inefficiency and redundancy in 4D LiDAR panoptic segmentation for autonomous driving applications, offering a novel geometric approach that is incremental over existing deep learning methods.
The authors tackled 4D LiDAR panoptic segmentation by proposing ICP-4D, a training-free framework that uses geometric priors and the Iterative Closest Point algorithm for instance association, achieving state-of-the-art performance on SemanticKITTI and panoptic nuScenes datasets without extra training or inputs.
Dominant paradigms for 4D LiDAR panoptic segmentation are usually required to train deep neural networks with large superimposed point clouds or design dedicated modules for instance association. However, these approaches perform redundant point processing and consequently become computationally expensive, yet still overlook the rich geometric priors inherently provided by raw point clouds. To this end, we introduce ICP-4D, a simple yet effective training-free framework that unifies spatial and temporal reasoning through geometric relations among instance-level point sets. Specifically, we apply the Iterative Closest Point (ICP) algorithm to directly associate temporally consistent instances by aligning the source and target point sets through the estimated transformation. To stabilize association under noisy instance predictions, we introduce a Sinkhorn-based soft matching. This exploits the underlying instance distribution to obtain accurate point-wise correspondences, resulting in robust geometric alignment. Furthermore, our carefully designed pipeline, which considers three instance types-static, dynamic, and missing-offers computational efficiency and occlusion-aware matching. Our extensive experiments across both SemanticKITTI and panoptic nuScenes demonstrate that our method consistently outperforms state-of-the-art approaches, even without additional training or extra point cloud inputs.