3D Extended Object Tracking based on Extruded B-Spline Side View Profiles

TL;DR

The paper addresses 3D extended object tracking for autonomous systems by representing a vehicle's 3D extent as an extrusion of a 2D side-view profile described by B-spline curves. A complete EKF-based tracker is derived, with a joint state including 3D kinematics and a parametric shape state defined by the B-spline control points. The measurement model uses a level-set formulation with boundary and cap points, and the approach is validated on CARLA simulations with simulated lidar/radar data and on a real Vod dataset. Results indicate robust motion tracking and competitive shape estimation, with performance improving as the number of spline control points increases and with radar providing favorable density in certain views.

Abstract

Object tracking is an essential task for autonomous systems. With the advancement of 3D sensors, these systems can better perceive their surroundings using effective 3D Extended Object Tracking (EOT) methods. Based on the observation that common road users are symmetrical on the right and left sides in the traveling direction, we focus on the side view profile of the object. In order to leverage of the development in 2D EOT and balance the number of parameters of a shape model in the tracking algorithms, we propose a method for 3D extended object tracking (EOT) by describing the side view profile of the object with B-spline curves and forming an extrusion to obtain a 3D extent. The use of B-spline curves exploits their flexible representation power by allowing the control points to move freely. The algorithm is developed into an Extended Kalman Filter (EKF). For a through evaluation of this method, we use simulated traffic scenario of different vehicle models and realworld open dataset containing both radar and lidar data.

Figures (7)

• Figure 1: The concept of extruding a side view profile into a 3D shape. The black contour lines represent the estimated model. The blue mesh is the true mesh of the object obtained in CARLA carla. The 3 axis represents the body frame of reference of the object. The red points defined in the body frame are the control points of the B-spline curve. They can be freely placed.
• Figure 2: The setup of the simulative environment for the evaluation. Four sets of sensors are placed at the position shown. Their field of view is represented by the colored area. A vehicle moves from the left to the right side. The driving behavior along the trajectory is marked in the figure.
• Figure 3: The vehicle models used in the simulation test.
• Figure 4: An example result of one frame. The red dots are the sampled point cloud as measurements.
• Figure 5: The result of state estimation for different vehicles. The results with different measurement data for all three models are shown. The shape estimation evaluation is shown with different numbers of control points of the B-spline curve.