Edge Detection Algorithm Used in a 3D Sketching & Photogrammetry System

Abstract

Abstract A clear 3D representation is vital to prototyping in design. 3D modeling is a popular approach to design structures in 3D, resulting in accurate models that can be viewed from multiple perspectives. However, 3D modeling is too costly and rigid for designers because it has a steep learning curve. Existing heavy-weight 3D software packages are usually not suitable for ideation. Therefore, we consider using sketches to convey ideas in 3D as a more natural way for prototyping. Even if most sketches are loose and 2D, people can still identify 3D structures from them. However, most sketching tools only support drawing in 2D, motivating us to find a solution to bridging 2D sketches and 3D structures. To fill the gap, we propose a 3D sketching system that facilitates sketch creation in 3D. Users can create a sketch in 3D and view it from multiple perspectives, without having to create multiple 2D sketches. This application also supports reconstructing a point cloud from images taken from existing structures. We designed a novel edge detection algorithm derived from the Hough transform to extract borders as strokes from a point cloud reconstructed from captured images. This includes fitting multiple planes to the point cloud and creating strokes by projecting points to the planes. We also present real-world use cases to evaluate our system. Our application allows designers to prototype more rapidly, opening a gate to more user-friendly ideation processes.