面向复杂场景的激光雷达地面分割算法

A lidar ground segmentation algorithm for complex scenes

针对现有地面分割方法在复杂场景下存在地面分割不准确、计算量大等问题,提出了一种基于先验信息采样一致性拟合和小型障碍物剔除的激光雷达地面分割方法。该方法首先利用激光雷达点云的先验信息指导数据采样并快速迭代估计地面模型,计算点云分布特征因数以评估地面模型,从而实现噪声环境下地面基准面的快速、准确拟合。随后,构建高度和法向量联合特征,剔除基准面上的小型障碍物,最终实现复杂场景下的地面准确分割。实验结果表明,该地面分离算法在复杂场景中具有较强的适应性和实时性,在保证分割效果的情况下较传统的算法效率提升30%。

The existing ground segmentation methods have problems of inaccurate ground segmentation and large amount of calculation in complex scenes. To address these issues, a lidar ground segmentation method is proposed, which is based on the prior information sampling consistent fitting and small obstacle removal. Firstly, the method uses the prior information of the lidar point cloud to guide data sampling. And the ground model is fast iteratively estimated. The point cloud distribution characteristic factors are used to evaluate the ground model. In this way, the fast and accurate fitting of the ground reference plane in the noise environment can be realized. Then, a joint feature of height and normal vector is established to eliminate small obstacle on the datum surface. Finally, the accurate ground segmentation in complex scenes can be achieved. Experimental results show that the ground separation algorithm has strong adaptability and real-time performance in complex scenes. The efficiency of the traditional algorithm is improved by 30% while ensuring the segmentation effect.