LiDAR-based 3D scene perception is a fundamental and important task for autonomous driving. Most state-of-the-art methods on LiDAR-based 3D recognition tasks focus on single frame 3D point cloud data, and the temporal information is ignored in those methods. We argue that the temporal information across the frames provides crucial knowledge for 3D scene perceptions, especially in the driving scenario. In this paper, we focus on spatial and temporal variations to better explore the temporal information across the 3D frames. We design a temporal variation-aware interpolation module and a temporal voxel-point refiner to capture the temporal variation in the 4D point cloud. The temporal variation-aware interpolation generates local features from the previous and current frames by capturing spatial coherence and temporal variation information. The temporal voxel-point refiner builds a temporal graph on the 3D point cloud sequences and captures the temporal variation with a graph convolution module. The temporal voxel-point refiner also transforms the coarse voxel-level predictions into fine point-level predictions. With our proposed modules, the new network TVSN achieves state-of-the-art performance on SemanticKITTI and SemantiPOSS. Specifically, our method achieves 52.5\% in mIoU (+5.5% against previous best approaches) on the multiple scan segmentation task on SemanticKITTI, and 63.0% on SemanticPOSS (+2.8% against previous best approaches).
翻译:以 LiDAR 为基础的 3D 场景感知是自动驱动的一项根本性重要任务。 在基于 LiDAR 的 3D 3D 身份识别任务上, 多数最先进的先进方法都以单一框架 3D 点云数据为重点, 而时间信息在这些方法中被忽略。 我们争辩说, 跨框架的时间信息为 3D 场景感知提供了至关重要的知识, 特别是在驱动情景中。 在本文中, 我们侧重于空间和时间变异, 以更好地探索三D 框架之间的时间信息。 我们设计了一个时间变异P 干涉模块和一个时间对口点修正器, 以捕捉取 4D 点云中的时间变异 。 时间变异觉间间对调通过捕捉空间一致性和时间变异信息, 从上和当前框架产生本地的特性。 我们的 SESNEO- 5 5 和 SEN-K 5 之前的SeNSO-% CSOIT 和之前的SEN- K- 5 格式, 通过我们的最佳模块, 在SESNS- 5 IM- 5 前的SEN- 5 IM- IM- 5 IM- 5 5 任务段上, 实现了我们以前的SET- 5 5 5 的SET- ta- ta- sem- ta- ta- ta- ta- ta- sem- sex- sex- ta- ta- sex- ta- ta- sal- ta- sal- sal- sal- pal- sal- sal- sal- sal- sal- sal 5 5 5 5 com- sal-s- sal- sal- sal- sal- ta- ta- 5 5 5 5 5 的方法, 在以前的模块中, 在以前的模块中, 在前的新的模块上, 5 com- sal- sal- sal- sal- sal- sal-s-s- sal- sal- 5 5 5 com-s-s-s-s-s- s- s- s- sal- sal- sal- sal-s- 5 5 5 5