Using Scanned Mesh Data for Auto-Digitized 3D Modeling: Abstract and Introductionby@rendering
162 reads

Using Scanned Mesh Data for Auto-Digitized 3D Modeling: Abstract and Introduction

tldt arrow

Too Long; Didn't Read

A paper regarding the automatic generation of accurate floor plans and 3D models from scanned mesh data for interior design and navigation.
featured image - Using Scanned Mesh Data for Auto-Digitized 3D Modeling: Abstract and Introduction
Rendering Technology Breakthroughs HackerNoon profile picture


(1) Ritesh Sharma, University Of California, Merced, USA [email protected];

(2) Eric Bier, Palo Alto Research Center, USA [email protected];

(3) Lester Nelson, Palo Alto Research Center, USA [email protected];

(4) Mahabir Bhandari, Oak Ridge National Laboratory, USA [email protected];

(5) Niraj Kunwar, Oak Ridge National Laboratory, USA [email protected].

Abstract and Intro

Related Work



Conclusion & Future work and References


This paper describes a novel approach for generating accurate floor plans and 3D models of building interiors using scanned mesh data. Unlike previous methods, which begin with a high resolution point cloud from a laser range-finder, our approach begins with triangle mesh data, as from a Microsoft HoloLens. It generates two types of floor plans, a "pen-and-ink" style that preserves details and a drafting-style that reduces clutter. It processes the 3D model for use in applications by aligning it with coordinate axes, annotating important objects, dividing it into stories, and removing the ceiling. Its performance is evaluated on commercial and residential buildings, with experiments to assess quality and dimensional accuracy. Our approach demonstrates promising potential for automatic digitization and orientation of scanned mesh data, enabling floor plan and 3D model generation in various applications such as navigation, interior design, furniture placement, facilities management, building construction, and HVAC design.

Keywords: Clustering based methods · Floor plans · Augmented Reality· 3D Models.

1 Introduction

Floor plans are useful for many applications including navigating in building interiors; remodeling; efficient placement of furniture; placement of pipes; heating, ventilation, and air conditioning (HVAC) design; and preparing an emergency evacuation plan. Depending on the application, different kinds of floor plan are appropriate. For remodeling building interiors or designing HVAC systems, users may prefer a drafting-style floor plan that focuses on planar walls and removes furniture and other clutter. For furniture placement, navigation, or evacuation planning, users may prefer a more detailed floor plan that shows the positions of furniture, cabinets, counter tops, etc. In either case, producing a floor plan can be time consuming, requiring expert skills, such as measuring distances and angles or entering data into a CAD program. Furthermore, it may need to be done more than once because a building changes when walls and furniture are moved, added, or removed. So it is valuable to be able to generate floor plans automatically with little or no training.

To generate floor plans, it helps to begin with accurate data that can be collected automatically. Laser range finders, smartphones, tablets, and augmented reality (AR) headsets are some of the devices that have made it easier to collect high-resolution building data in the form of RGBD images, point clouds, and triangle meshes. In this paper, we describe a method for generating drafting style and pen-and-ink-style floor plans by leveraging incomplete and imperfect triangle mesh data. This approach efficiently generates both types of floor plans accurately, supporting a wide range of applications.

Main Contribution We describe a new method for generating accurate floor plans using poorly captured triangle mesh data from the Microsoft HoloLens 2. The main contributions are:

– A modified Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, using blocks to capture wall height and thickness.

– An orientation-based clustering method that finds walls at arbitrary angles.

– The use of k-means clustering to rotate the mesh to the principal axes and to identify the floor and ceiling.

– Generating two kinds of precise floor plans from incomplete mesh data.

This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license.