A view-tree method to compute viewsheds from digital elevation models

Viewshed computation is a central component for visibility analysis. The majority of existing viewshed computation methods use regular square grid digital elevation models (DEMs) and have limitations in considering spatial relationships among observers, targets, and line-of-sight (LOS). Therefore, considering that the visibility is dominated by LOS between observers and targets, this study proposes a method to compute viewsheds using a new type of tree structure, the “view-tree,” that can efficiently extract LOS from DEMs for viewshed computation. The proposed method first constructs a view-tree using cells in regular square grid DEM according to the spatial occlusion relationships among cells. Then, the method traverses the view-tree to judge the visibility of each tree node and derives the viewshed. The results show that the view-tree method is 40% faster than the traditional R2 method. The view-tree method also reduces the error and omission rates by approximately 60% compared with the popular XDraw method.

视域计算是可视性分析的核心组成部分。现有绝大多数视域计算方法均采用规则正方形格网数字高程模型（Digital Elevation Model, DEM），但在考量观测者、目标物与视线（Line-of-Sight, LOS）之间的空间关联方面存在局限。鉴于可视性由观测者与目标物之间的视线关系主导，本研究提出一种基于新型树结构——“视域树（view-tree）”的视域计算方法，可高效从数字高程模型中提取视线以完成视域计算。所提方法首先依据规则正方形格网数字高程模型中格网单元间的空间遮挡关系构建视域树，随后通过遍历该视域树以判断各树节点的可视性，进而推导得到视域范围。实验结果表明，视域树方法的运行速度较传统R2方法提升40%；相较于主流的XDraw方法，该方法的误差率与漏检率也分别降低约60%。