汽车工程 ›› 2024, Vol. 46 ›› Issue (4): 605-616.doi: 10.19562/j.chinasae.qcgc.2024.04.006

• • 上一篇    

基于扩展卡尔曼滤波与机器视觉融合的道路侧向坡度估计

严运兵(),岳铭浩,李海玮   

  1. 武汉科技大学汽车与交通工程学院,武汉 430065
  • 收稿日期:2023-09-14 修回日期:2023-10-29 出版日期:2024-04-25 发布日期:2024-04-24
  • 通讯作者: 严运兵 E-mail:yyb@wust.edu.cn
  • 基金资助:
    国家自然科学基金面上项目(51975428)

Estimation of Road Side Slope Based on Fusion of Extended Kalman Filter and Machine Vision

Yunbing Yan(),Minghao Yue,Haiwei Li   

  1. School of Automobile and Traffic Engineering,Wuhan University of Science and Technology,Wuhan 430065
  • Received:2023-09-14 Revised:2023-10-29 Online:2024-04-25 Published:2024-04-24
  • Contact: Yunbing Yan E-mail:yyb@wust.edu.cn

摘要:

为解决现有算法难以准确估计前方道路侧向坡度的问题,提出了一种基于扩展卡尔曼滤波(EKF)与机器视觉(VB)融合的道路侧向坡度估计方法。首先,建立含有侧向坡度的车辆2自由度模型,通过EKF估计出侧向坡度与车辆侧倾角的叠加态,由侧向加速度乘以适当增益解耦出车辆侧倾角,得到EKF道路侧向坡度估计值;其次,通过视觉成像原理分析二维图像中道路侧向坡度与图像中相关参数的几何关系,得到VB道路侧向坡度估计值;最后,通过数据融合得到最终的道路侧向坡度估计值,使估计结果冗余互补。仿真和实车试验结果表明,该融合算法能够适用于道路侧向坡度变化的坡道,并显著提高了估计精度。

关键词: 侧向坡度, 坡度估计, 扩展卡尔曼滤波, 机器视觉, 数据融合

Abstract:

To address the difficulty of accurately estimating the lateral slope of the road ahead using existing algorithms, a road lateral slope estimation method based on the fusion of Extended Kalman Filter (EKF) and Machine Vision Based (VB) is proposed. Firstly, a two-degree of freedom model for vehicles with lateral slope is established, and the superposition state of lateral slope and vehicle roll angle is estimated through EKF, with the vehicle roll angle decoupled by multiplying the lateral acceleration with an appropriate gain to obtain the estimated lateral slope of the EKF road. Secondly, based on the principle of visual imaging, the geometric relationship between the lateral slope of roads in two-dimensional images and the relevant parameters in the images is analyzed to obtain an estimated value of the lateral slope of roads in VB. Finally, the final estimation of road lateral slope is obtained through data fusion, making the estimation results redundant and complementary. The simulation and actual vehicle test results show that the fusion algorithm can be applied to slopes with varying lateral slopes and significantly improve the estimation accuracy.

Key words: lateral slope, slope estimation, extended Kalman filter, machine vision, data fusion