Scribd
Upload
82 views2 pages

BundlerPgP Instructions

The document provides instructions for using Bundler Photogrammetry Package software to perform structure-from-motion (SfM) and multi-view stereo (MVS) photogrammetric reconstruction from images. It describes installing the software, preparing image data, running SfM to generate a sparse point cloud, and running MVS to generate a dense point cloud. The results can be visualized using free software like Meshlab or CloudCompare.

Uploaded by

Kenneth_Alba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
82 views2 pages

BundlerPgP Instructions

The document provides instructions for using Bundler Photogrammetry Package software to perform structure-from-motion (SfM) and multi-view stereo (MVS) photogrammetric reconstruction from images. It describes installing the software, preparing image data, running SfM to generate a sparse point cloud, and running MVS to generate a dense point cloud. The results can be visualized using free software like Meshlab or CloudCompare.

Uploaded by

Kenneth_Alba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Using ‘Bundler Photogrammetry Package’ for SfM-MVS reconstructions

1. Setup
Hardware and software requirements:
Windows operating system, 64-bit computer, Nvidia graphics card.
Bundler Photogrammetry Package: download from http://files.neonascent.net/SFM.zip

Software installation
Extract SFM.zip to C:\
The exact path is important and others will not work. Ensure that the extraction has given you folders:
C:\SFM\ bin C:\SFM\ bundler_doc
C:\SFM\ bin64 C:\SFM\ cygwin C:\SFM\ examples

2. Carry out a reconstruction


Data preparation
1) Make a new folder inside C:\SFM\examples\ to hold your reconstruction project and copy in your
image files (.jpg format) [NOTE – do not use spaces in your folder or image names]
2) Copy MakeList.bat from C:\SFM\ into your project folder and run the file (double-click). This
creates a single file list.txt that contains a list of the images to be processed.

Structure-from-motion (SfM) sparse reconstruction


3) Copy sparseRecon64.bat from C:\SFM\ into your project folder and run it (double-click).
This runs the SfM script – follow the instruction to hit a key to start. The first step is the SIFT
feature extraction and you should see text similar to :
[Extracting Sift feature : 33%] – (1/9, 1807 features)
If there is an error, this is likely to be to do with your graphics card (SIFT processing is carried out
on the graphics card) – a remedy for which is out of scope of these instructions. When SIFT
feature extraction and initial matching has completed, you need to hit a key to start Bundler.

Bundler requires the sensor width of the camera to initialize processing. In most cases Bundler extracts this
from either the EXIF header data of the images, or from a text file used by the software. However, if the
data are not found, you will see
[Couldn't find CCD width for camera ...]
and you need to add the details yourself to the text file. To do this :
1) Find the CCD width of your camera in mm; values are usually given on manufacturer’s website or
at camera review sites such as DPReview (http://www.dpreview.com/).
2) You also need to determine the text in your image files that Bundler uses to identify which camera
was used. In Windows Explorer, right-click on an image and go to Properties. Select the Details or
Summary tab (clicking ‘Advanced’if necessary), and find the value for the Camera Model property.
3) Using a text editor, insert a line containing you camera details to C:\SFM\bin\extract_focal.pl file in
the same format as the lines for other camera data:
%ccd_widths = (
…
"YOUR CAMERA MODEL" => ww,
…

);
where ww is the CCD width in mm. Note that anything written after a ‘#’ character is a comment
only and not used by the software. Save the file and retry sparseRecon64.bat. If the problem
remains, double check that you have exactly the same text in your entry to the CCD list as in the
image properties (including capitalization and spaces).

When Bundler is running then a sub-directory called bundle will be produced in your project
folder. As the reconstruction proceeds, Bundler writes incremental output files in the form
bundle_nnn.out and pointsnnn.ply, where nnn is the number of images that have been
successfully incorporated into the model so far. The .ply files are ASCII format 3D point clouds
which show the positions of the reconstructed cameras and the 3D surface. They can be viewed
with a variety of software (see below).

Multi-view stereo (MVS) dense reconstruction


Once Bundler has finished, the SfM step is complete and the MVS dense reconstruction can be started:

4) Copy denseRecon_batch.bat and denseRecon.vbs from C:\SFM\ into your project folder
and run denseRecon.vbs
5) Enter the values for the match parameters - the recommended ones usually give good result but
the panel below explains more about the individual parameters

Images per segment: This sets the maximum number of images that will be clustered together (by
CMVS) for simultaneous matching. More is better, but the maximum number is limited by RAM
and will vary with the density of the reconstruction. Fewer means that more image clusters will be
produced for matching one after another.
Reconstruction level and Voxel size: Both these parameters effectively vary the density at which the
matching is carried out. Denser matching require more RAM and is much slower than less dense
matching. The details are not straightforward, but level determines the number of times images are
decimated before matching (i.e. level = 1 (recommended) represents images used at half-
resolution). Voxel size determines how often a match is attempted – i.e. every nth pixel in the x and
y directions of the sampled images. For the recommended voxel size = 2, then a match is attempted
for every fourth pixel in an image.
Reconstruction threshold: A threshold that is related to correlation values in the matching process and
is used to filter out bad matches. Larger (up to 1.0) values mean fewer but more reliable points,
smaller values retain more points, but the quality can be lower. A value of 0.9 is recommended.

The matching process may take some time. At the end there should be one or more .ply files within the
subdirectory \pmvs\models of your project folder.

3. Visualise the results


All .ply files produced are in ASCII format and contain the point coordinates and color information after
number of header lines. ASCII .ply files can be read directly by most point cloud processing software.
Good freeware applications are:

Meshlab: http://meshlab.sourceforge.net/

Cloudcompare http://www.danielgm.net/cc/

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy