【摘要】 自动驾驶汽车的凸凹不平道路和异常路面行驶，不仅要考虑道路曲率等因素，还需要对路面凸起、凹坑等特征和病害进行识别建模，提高车辆通过性、安全性和舒适性。为此，本文提出了一种基于激光雷达的凸凹不平道路几何参数识别和模型重构方法。首先，将局部加权散点平滑方法（Lowess）首次应用于激光雷达点云处理，提高了激光雷达点云数据的平滑性；其次，提出了基于斜率阈值分割的路面几何参数识别方法，通过设置斜率阈值对道路凸起与凹坑进行识别提取；再次，通过识别特征点云边界构建了带约束的分段多项式函数路面连续典型特征拟合数学模型。最后，通过建立的室内路面典型特征沙盘模型及路面实测数据，应用本文提出的方法，对凸凹不平道路的凹坑和凸起等特征进行了识别和模型重构。结果表明，分段多项式拟合方法在拟合次数5～6次时达到拟合效果极限位置，此时各个场景中92%的数据点拟合均方根误差在0～0.015 m范围内，本文提出的方法能准确完成凸凹不平道路几何参数识别，实现路面典型特征三维数学模型重构。更多还原

【Abstract】 For autonomous vehicles driving on uneven road and abnormal road, not only factors such as road curvature need to be considered, but also characteristics and diseases such as road bumps and potholes need to be identified and modeled, so as to improve vehicle passability, safety and comfort. Therefore, this paper proposes a surface geometry parameter recognition and model reconstruction method of uneven road based on the Lidar.Firstly, the Locally Weighted Scatterplot Smoothing(Lowess) method is applied to the lidar point cloud processing for the first time to improve the smoothness of the lidar point cloud data. Secondly, a surface geometry parameter recognition method based on slope threshold segmentation is proposed to identify and extract the road bumps and pits by setting slope threshold. Thirdly, the 3D mathematical model based on piecewise polynomial function with constraints for road surface geometry parameter of uneven road is built. Finally, through the indoor sand table model of typical characteristics of the road and the measured data, the road geometry parameters’ recognition and model reconstruction are carried out by the proposed method. The results show that the piecewise polynomial fitting method achieves the best fitting effect when the fitting times are 5-6 times, and the root mean square error of 92% data points in each scene is within the range of 0～0.015m. The proposed method can accurately achieve the surface geometric parameter identification of uneven road and realize the 3D mathematical model reconstruction of the typical geometry feature.更多还原