Advanced

Automatic Room-Blueprint Extraction using SfM

Ericson, Joakim LU and Jardby, Filip LU (2020) In Master’s Theses in Mathematical Sciences FMAM05 20201
Mathematics (Faculty of Engineering)
Abstract
In this thesis the possibility of automatically generating a blueprint of an indoor area, for example a grocery store, using a smartphone camera and images from existing analytic or surveillance cameras was examined. The method used for the transition between 2D camera images and 3D reconstruction was Structure from Motion. Besides the blueprint, two other key features was implemented. One of these key features was to allow the user to position new cameras into the indoor area using a 3D GUI and to visualize what these cameras can see, given their rotation. The camera's field of view was also visualized in the 2D blueprint for improved overview. The other key feature was the ability to make an "update check", where an image from the active... (More)
In this thesis the possibility of automatically generating a blueprint of an indoor area, for example a grocery store, using a smartphone camera and images from existing analytic or surveillance cameras was examined. The method used for the transition between 2D camera images and 3D reconstruction was Structure from Motion. Besides the blueprint, two other key features was implemented. One of these key features was to allow the user to position new cameras into the indoor area using a 3D GUI and to visualize what these cameras can see, given their rotation. The camera's field of view was also visualized in the 2D blueprint for improved overview. The other key feature was the ability to make an "update check", where an image from the active analytic or surveillance camera is compared to a reference image (previously captured by the camera) in order to detect missing or added objects. (Less)
Popular Abstract
Obtaining a 3D model of a room or facility is today a rather expensive process due to the expensive camera equipment required. In this thesis it has been shown that it is possible to obtain a precise 3D model of an indoor space, using nothing but a smartphone for capturing images and a computer for making the calculations required.

This thesis focuses on the use case of where a 3D model of a grocery store is constructed. The two major reasons for specifically choosing a grocery store were first, that the majority of grocery stores already have an integrated camera surveillance system which makes the process of live updating changes to the 3D model possible and secondly because it is terribly difficult to manually measure up a grocery... (More)
Obtaining a 3D model of a room or facility is today a rather expensive process due to the expensive camera equipment required. In this thesis it has been shown that it is possible to obtain a precise 3D model of an indoor space, using nothing but a smartphone for capturing images and a computer for making the calculations required.

This thesis focuses on the use case of where a 3D model of a grocery store is constructed. The two major reasons for specifically choosing a grocery store were first, that the majority of grocery stores already have an integrated camera surveillance system which makes the process of live updating changes to the 3D model possible and secondly because it is terribly difficult to manually measure up a grocery store due to its large number of alleys.

The idea is that the intended user has two choices; either go around the store capturing a picture every meter or to simply videotape the store. Capturing images will lead to a slightly more precise 3D model, but the process will become more tedious than videotaping. Afterwards the pictures or videotape will be uploaded to a computer and the calculations performed. The user will then be able to run a GUI program with the options of either displaying the store in 2D or 3D. The program offers several usable features such as 2D and 3D measurements, view points of hypothetically placed surveillance cameras, item removal check (if the user has access to live surveillance footage) and a good 3D overview for floor planning.

By succeeding in constructing a prototype program like this, the potential of SfM (structure from motion) has been proven for commercial use. This technique could be used for any purpose where it would be beneficial to know the exact geometry of a room. Another area of use could be where homeowners measure their homes in order to virtually plan their next kitchen, which furniture to fit where or simply just to receive a blueprint over their house. (Less)
Please use this url to cite or link to this publication:
author
Ericson, Joakim LU and Jardby, Filip LU
supervisor
organization
alternative title
Automatisk generering av planritningar från Struktur-Rörelse skattningar
course
FMAM05 20201
year
type
H2 - Master's Degree (Two Years)
subject
publication/series
Master’s Theses in Mathematical Sciences
report number
LUFTMA-3409-2020
ISSN
1404-6342
other publication id
2020:E28
language
English
id
9014877
date added to LUP
2020-06-24 13:42:23
date last changed
2020-06-24 13:42:23
@misc{9014877,
  abstract     = {In this thesis the possibility of automatically generating a blueprint of an indoor area, for example a grocery store, using a smartphone camera and images from existing analytic or surveillance cameras was examined. The method used for the transition between 2D camera images and 3D reconstruction was Structure from Motion. Besides the blueprint, two other key features was implemented. One of these key features was to allow the user to position new cameras into the indoor area using a 3D GUI and to visualize what these cameras can see, given their rotation. The camera's field of view was also visualized in the 2D blueprint for improved overview. The other key feature was the ability to make an "update check", where an image from the active analytic or surveillance camera is compared to a reference image (previously captured by the camera) in order to detect missing or added objects.},
  author       = {Ericson, Joakim and Jardby, Filip},
  issn         = {1404-6342},
  language     = {eng},
  note         = {Student Paper},
  series       = {Master’s Theses in Mathematical Sciences},
  title        = {Automatic Room-Blueprint Extraction using SfM},
  year         = {2020},
}