Enhanced Temporal Data Organization for LiDAR Data in Autonomous Driving Environments

Abstract

One of the most important tasks for autonomous cars is the perception of the environment. In particular, the detection and tracking of objects is vital for further applications. We present a new real-time method to organize point cloud data provided by a LiDAR sensor. The main contribution of this method is the linking of 3D points from different time frames. With this connection, it is possible to traverse through the data over time. In addition, an efficient 2D data organization allows fast access to neighboring information of the 3D data. This makes it very suitable for tasks like model creation and clustering. Based on the obtained spatial and temporal neighboring information, tasks such as object detection, tracking and prediction can be solved directly.