Mono-Camera based 3D Object Tracking Strategy for Autonomous Vehicles
Mono-Camera based 3D Object Tracking Strategy for Autonomous Vehicles
复制标题
基于单摄像头的自动驾驶车辆 3D 对象跟踪策略
DOI：
10.1109/ivs.2018.8500482
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
N. Suganuma
中科院分区：
文献类型：
--
作者：
Akisue Kuramoto;Mohammad Aldibaja;R. Yanase;Junya Kameyama;Keisuke Yoneda;N. Suganuma
本文提出了一种方法来计算远检测车辆的三维位置。主要是，在自动驾驶过程中，必须精确估计车辆的距离，以制定安全的路径规划。创建3D相机模型以将像素位置映射到相对于车辆平面的距离值和失真参数。为了提高距离精度，扩展卡尔曼滤波（EKF）框架被设计用于跟踪检测到的车辆之间的导数关系的基础上的相机和世界坐标系。实验结果表明，所提出的方法是能够成功地跟踪三维位置与足够的精度相比，激光雷达和雷达跟踪系统的成本和稳定性。
This paper proposes an approach to calculate 3D positions of far detected vehicles. Mainly, the distance from the vehicles during autonomous driving must be estimated precisely to strategize a safe path planning. A 3D camera model is created to map the pixel positions to the distance values with respect to the vehicle plane and the distortion parameters. In order to refine the distance accuracy, the Extended Kalman Filter (EKF) framework is designed to track the detected vehicles based on the derivative relationship between the camera and world coordinate systems. The experimental results indicate that the proposed method is capable to successfully track 3D positions with sufficient accuracy compared to LIDAR and Radar based tracking systems in terms of cost and stability.