Modelling accurate ground surfaces in urban areas is important for surveying and mapping. Breaklines along roads are critical for both digital elevation models and high precision maps. This article presents a new breakline-preserving ground interpolation method for point clouds acquired by mobile laser scanning (MLS). The proposed method needs only point coordinates as input. It first initializes unknown regions by the nearest neighbouring interpolation then matched patches in known regions are found along edges. Next, Poisson interpolation is utilized to improve the elevation accuracy. An edge-guided patch regularization method is applied before patch blending to remove noise and improve gradient accuracy. The proposed method was tested on two datasets acquired by different MLS systems. This paper quantitatively evaluates the proposed method as well as the existing solution in two samples of complete points. These results demonstrate that the proposed method outperforms the existing method in terms of visual coherence and mean absolute differences. Moreover, the edge information which is self-contained in point clouds has been proven useful in ground interpolation.

在城市地区对准确的地面进行建模对于测量和制图很重要。沿道路的折线对于数字高程模型和高精度地图都是至关重要的。本文为通过移动激光扫描（MLS）获取的点云提供了一种新的保留折线的地面插值方法。所提出的方法仅需要点坐标作为输入。它首先通过最近的相邻插值初始化未知区域，然后沿着边缘找到已知区域中的匹配色块。接下来，利用泊松插值来提高仰角精度。在面片融合之前应用边缘引导的面片正则化方法以消除噪声并提高梯度精度。在不同的MLS系统获取的两个数据集上测试了该方法。本文定量评估了提出的方法以及两个完整点样本中的现有解决方案。这些结果表明，在视觉连贯性和平均绝对差异方面，该方法优于现有方法。此外，已经证明，点云中包含的边缘信息可用于地面插值。