Structure from Motion Photogrammetry: Constructing 3 Dimensional Structure from Ordinary Photography

Project ID: 3835
Principal Investigator: Eric Peterson
Research Topic: Water Resource Data Analysis
Funded Fiscal Years: 2015
Keywords: photogrammetry, structure from motion, sfm, computer vision, camera, three dimensional data, point cloud, terrain, topography, survey

Research Question

We seek to maximize Reclamation's capabilities in a developing field of 3D data acquisition referred to as Structure from Motion (SfM). Acquiring 3D data is increasingly vital to Reclamation for its use in numerous fields, eg:
• CAD design of structures and restoration features
• monitoring water resources
• modeling for decision support of management actions:
o computer modeling such as SRH-2D hydraulic models
o physical modeling in flumes, coupled with 3D printing

Related to traditional photogrammetry, SfM uses algorthims to automatically identify and triangulate objects across multiple photographs from different points of view. Computerized automation extends SfM to 3D reconstruction of point clouds for entire scenes and recently has achieved sufficient speeds for practical spatial reconstruction over large areas with numerous photographs yielding millions of data points.

Pilot Tests by the PI reconstructed river banks in much greater detail than conventional aerial lidar from a boat-based camera, though positional accuracy was somewhat greater than aerial lidar. Accuracy might be improved with different camera lenses, photographic techniques, and frame rates.

In closer and more controlled situations, accuracy could be almost unlimited; SfM can reconstruct small objects with sub-millimeter accuracy.

Proper application of SfM also requires knowledge of potential pitfalls. For example, shrubs on river banks can form intricate texturing that is difficult to match between photographs for triangulation by SfM algorithms.

Several Reclamation offices already use SfM for projects ranging from dam removal to mapping material surface deterioration. While SfM has great potential, methods can be so flexible and variable that it can be difficult to identify the optimal method for a particular need. Exchanging experience from others should enhance Reclamation's use of SfM.

Need and Benefit

SfM provides a flexible and convenient alternative to various laser scanning technologies, thereby expanding Reclamation's toolbox for 3D data acquisition. Initial data collection requires only a standard camera – even most cell phones are sufficient. Thus it can be used in places or conditions where laser scanning is difficult. Furthermore, SfM can be used to address impromptu needs when a laser scanner may not be quickly accessible. Lastly, SfM is a low-cost option enabling more frequent spatial monitoring of Reclamation's actions including before and after comparisons, or even topographic evolution at numerous points through time. Should Unmanned Aerial Vehicle technologies become more accessible, inexpensive cameras can be easily attached to UAVs for developing topographic models with SfM.

The more Reclamation's capabilities with SfM improve, the better it will be able to take advantage of this cost-efficient, time-efficient, and flexible technology.

Contributing Partners

Contact the Principal Investigator for information about partners.

Research Products

Bureau of Reclamation Review

The following documents were reviewed by experts in fields relating to this project's study and findings. The results were determined to be achieved using valid means.

Structure from Motion Photogrammetry: Constructing 3 Dimensional Structure from Ordinary Photography (final, PDF, 4.5MB)
By Eric Peterson
Publication completed on September 30, 2015

This paper documents the learning of a ''community of interest'' seeking to maximize Reclamation's capabilities in a developing field of 3D data acquisition referred to as Structure from Motion (SfM) Photogrammetry. Photogrammetry is the use of two dimensional (2D) images to provide measurement data. Structure-from-motion refers to a set of algorithms from computer vision sciences that assist photogrammetry by automatically detecting and matching features across multiple images, then triangulating positions to form 3D point clouds. SfM Photogrammety data collection (photography) and processing are discussed with a focus on PhotoScan Professional software, but with acknowledgement of no-cost software options where applicable. Techniques and equipment for optimizing efficiency and accuracy are included. Several test cases are demonstrated, both to highlight advantages to SfM Photogrammetry and to provide awareness of challenges. A table summarizing relative advantages of photogrammetry versus laser scanning technologies is provided.

Return to Research Projects

Last Updated: 6/22/20