| 基于双目视觉计算的车辆跟驰状态实时感知系统 |
| |
| 引用本文： | 王昊,刘雍翡.基于双目视觉计算的车辆跟驰状态实时感知系统[J].中国公路学报,2019,32(12):88. |
| |
| 作者姓名： | 王昊 刘雍翡 |
| |
| 作者单位： | 1. 东南大学 交通学院, 江苏 南京 211189;2. 东南大学 城市智能交通江苏省重点实验室, 江苏 南京 211189;3. 东南大学 现代城市交通技术江苏高校协同创新中心, 江苏 南京 211189 |
| |
| 基金项目： | 国家自然科学基金项目（51878161） |
| |
| 摘 要： | 双目视觉技术能够实现目标的识别与距离计算，在自动驾驶领域有很大的应用空间。然而，现阶段双目视觉存在光照干扰、遮挡、弱纹理区域歧义匹配等问题，影响其测量的准确性和可靠性。提出基于双目视觉的跟驰状态实时感知系统，该系统采用基于车辆跟驰模型的扩展卡尔曼滤波方法对车辆跟驰状态进行实时估计，包括跟驰距离、前后车速度差等。通过实际道路试验，证明了该系统能够识别并修正测量数据中的异常值，解决弱纹理区域误匹配问题。试验结果表明：25 mm焦距与12 mm焦距的双目系统跟驰间距测量值的平均误差分别为2.66%与9.14%；在相对速度测量方面，2种焦距系统的测量精度基本相同，平均误差均为1 m·s-1左右。所提出的方法在自动驾驶车辆环境感知领域有较好的应用前景。
|
| 关 键 词： | 交通工程 跟驰状态估计 双目视觉 扩展卡尔曼滤波 自动驾驶汽车 |
| 收稿时间： | 2019-03-20 |
Car-following State Real-time Estimation System Based on Binocular Vision |
| |
| WANG Hao,LIU Yong-fei.Car-following State Real-time Estimation System Based on Binocular Vision[J].China Journal of Highway and Transport,2019,32(12):88. |
| |
| Authors: | WANG Hao LIU Yong-fei |
| |
| Affiliation: | 1. School of Transportation, Southeast University, Nanjing 211189, Jiangsu, China;2. Jiangsu Key Laboratory of Urban ITS, Southeast University, Nanjing 211189, Jiangsu, China;3. Jiangsu Province Collaborative Innovation Center of Modern Urban Traffic Technologies, Southeast University, Nanjing 211189, Jiangsu, China |
| |
| Abstract: | Binocular vision has advantages in autonomous driving because it can detect objects and calculate distance simultaneously. However, there are still many unsolved problems in the application of binocular vision in real traffic conditions, such as illumination interference, occlusion, mismatching in weak texture regions, etc. The authors propose a real-time perception system of the car-following state based on binocular vision in this paper. The system applied an extended Kalman filter based on a car-following control model to estimate the car-following state in real time. The system can collect real-time data on the car-following state, including car-following distance, relative speed, etc. By analyzing the data collected from real-world road experiments, it was proved that application of the extended Kalman filter can accurately identify and correct outliers, solving the problem of mismatched weak texture regions. The experimental data show that the mean absolute percentage error of car-following distance measured by 25 mm and 12 mm cameras is 2.66% and 9.14%, respectively. The relative velocity measurement accuracy of the two kinds of cameras is similar. The mean absolute error of relative velocity is approximately 1 m·s-1. The proposed car-following state real-time estimation system has potential applications in the field of perception for autonomous vehicles. |
| |
| Keywords: | traffic engineering car-following state estimation binocular vision extended Kalman filter autonomous vehicle |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《中国公路学报》浏览原始摘要信息 |
|
点击此处可从《中国公路学报》下载全文 |
|
