This paper presents an effective and semi-automated method for detecting 3D edges in 3D point clouds with the help of high-resolution digital images. The effort aims to contribute towards addressing the unsolved problem of automated production of vector drawings from 3D point clouds of cultural heritage objects. Edges are the simplest primitives to detect in an unorganized point cloud and an algorithm was developed to perform this task. The provided edges are defined and measured on 2D digital images of known orientation, and the algorithm determines the plane defined by the edge on the image and its perspective center. This is accomplished by applying suitable transformations to the image coordinates of the edge points based on the Analytical Geometry relationships and properties of planes in 3D space. This plane inevitably contains the 3D points of the edge in the point cloud. The algorithm then detects and isolates those points which define the edge in the world system. Finally, the goal is to reliably locate the points that describe the desired edge in their true position in the geodetic space, using several constraints. The algorithm is firstly investigated theoretically for its efficiency using simulation data and then assessed under real conditions and under different image orientations and lengths of the edge on the image. The results are presented and evaluated.

中文翻译：

---

使用数字图像的3D点云边缘检测

本文提出了一种借助高分辨率数字图像检测3D点云中3D边缘的有效且半自动化的方法。这项工作旨在解决尚未解决的问题，即从文化遗产对象的3D点云自动生成矢量绘图。边缘是在无组织的点云中检测到的最简单的原语，因此开发了一种算法来执行此任务。所提供的边缘是在已知方向的2D数字图像上定义和测量的，并且算法确定图像上的边缘及其透视中心确定的平面。这是通过基于3D空间中的解析几何关系和平面属性将适当的变换应用于边缘点的图像坐标来实现的。该平面不可避免地包含点云中边缘的3D点。然后，该算法将检测并隔离那些定义世界系统中边缘的点。最后，目标是使用几个约束条件，将描述所需边缘的点可靠地定位在大地测量空间中的真实位置。首先使用仿真数据从理论上研究该算法的效率，然后在实际条件下以及在不同的图像方向和图像边缘的长度下对其进行评估。显示结果并进行评估。使用几个约束。首先使用仿真数据从理论上研究该算法的效率，然后在实际条件下以及在不同的图像方向和图像边缘的长度下对其进行评估。显示结果并进行评估。使用几个约束。首先使用仿真数据从理论上研究该算法的效率，然后在实际条件下以及在不同的图像方向和图像边缘的长度下对其进行评估。显示结果并进行评估。