Unit 1-SfM: Introduction to SfM
This unit introduces students to Structure from Motion (SfM). SfM is a photogrammetric technique that uses overlapping images to construct a 3D model of the scene and has widespread research applications in geodesy, geomorphology, structural geology, and other subfields of geology. SfM can be collected from a hand-held camera or an airborne platform such as an aircraft, tethered balloon, kite, or UAS (unmanned aerial system). After an introduction to the basics of SfM, students will design and conduct their own survey of a geologic feature, followed by an optional (but highly encouraged) introductory exploration of SfM data after returning from the field.
Unit 1-SfM Learning Outcomes
Students are able to:
- Design and conduct a simple SfM survey including:
- Make necessary calculations to determine the optimal survey parameters and survey design based on site and available time
- List the individual survey steps (workflow)
- List the equipment needed for a survey
- Make supporting field notes such as outcrop sketches, strike/dip measurements, and observations of outcrop conditions
- Process photographs taken in the field to generate a three-dimensional model of a geologic feature.
Unit 1-SfM Teaching Objectives
- Cognitive: Facilitate design and implement an SfM survey
- Behavioral: Promote student ability collecting SfM data using one or more platforms (ex., balloon, pole, UAS), set up related equipment (ex., targets and GPS), and make supporting field observations.
Context for Use
The content in Unit 1-SfM was designed for upper-level geoscience majors in a field geology course. The material works well for a group of approximately twenty (or fewer) students with an instructor and teaching assistant/s. If teaching a larger group of students, this can be scaled up with more survey setups, such as multiple handheld cameras or a combination of survey platforms. Fewer students in the course allows students more hands-on time operating the collection platform. If there are more than fifteen students, consider using two locations and two collection platforms, to enable all students to be involved in the data collection process. Concurrently doing Unit 1-SfM with Unit 1-TLS is another way to keep a larger group of students occupied. This content may also be modified to work in a classroom setting as a lecture (introduction to SfM) and lab (conducting a small survey of a feature, geologic or something available on campus). SfM requires a several hours to a day for post-collection processing before data exploration is possible, so it adapts well to interleaving with other activities or to an academic-year course. Canned data is also available if a collection platform or appropriate outcrop is unavailable. The unit may also be adapted to a one-day field trip for a geomorphology, structural geology, geophysics, remote sensing/GIS, or field methods class. Student experience with field observations, field maps, and trigonometry, in addition to other basic calculation skills is expected. In a field course, this unit is ideally situated mid-way through the course, as students will already have some field experience. As this is the introductory unit for the module, it pairs well with any of subsequent units. It is expected that most instructors will not use all the module's units, but combine Unit 1-SfM with 1–2 intermediate units and Unit 5 (Summative Assessment) and perhaps even Unit 1-TLS.
Description and Teaching Materials
1) Classroom introduction
This unit begins in a classroom setting, with a lecture presentation followed by the distribution of the SfM manuals and assignment packets. The lecture includes background on why geodetic surveying techniques are used in the earth sciences, and how SfM works. Similar material is provided in the manual. This segment should be brief (thirty minutes to one hour) to allow time for SfM surveying in the field. If possible, have another instructor photograph a small portion of the classroom while giving the first lecture so students can view the data. Roughly fifteen photos is ideal, as this number allows for quick data processing to generate a model of part of the room. These photos may be collected and processed prior to the lecture, to ensure the model is done processing. Take a break between the two presentations to show the students the equipment and have them practice setting up targets to break up the lecture time. Inventory the equipment to ensure that all necessary components are present; while this is happening, students may fill out the blank equipment list in their assignment packets. Students should bring all typical field equipment with them (writing utensils, straight edge/ruler, field notebook, AND calculator*).
*a calculator is essential to complete their work!
2) Field deployment and platform/survey overview
Introduce the field site, facilitating student discussion to determine the project objectives. Then introduce the platform and field techniques. This includes how to set up the platform of choice, the photo collection method (i.e., continuous video, automatic timer for photos, or remote controlled [the last two are recommended]), and the targets, as well as identifying the feature/s of geologic interest to be surveyed. Design the survey as a group, soliciting student input for the location of targets, GPS, and camera locations/collection path.
Note: GPS is not necessary for all surveys; scale bars or referencing an ALS or TLS point cloud of the same area for ground control points may be appropriate for the feature you are surveying. Refer to the SfM Guide for Instructors and Investigators to determine which methodology of scaling the data is appropriate for the feature of interest.
3) Conducting the survey
Students can be broken into teams of three to five people to set up the targets and GPS. The steps for conducting the survey vary based on the platform used for collecting photographs. After this, determine which students will be collecting which section of photographs. If just using a handheld camera or pole platform (ex., to collect data of a road cut or a planar outcrop with low dip), the camera locations should be planned by the group and a few photographers selected to photograph small sections. If using a balloon, use a two-pilot setup so pairs of students may rotate in. If using an Unmanned Aircraft System (UAS), rotate between students to pilot the device.* Aerial platforms should be used for larger-scale features, while pole platforms and handheld cameras can be used for outcrop-scale features. More students will be able to actively participate in the surveying process either: (A) using aerial platforms because a greater area will be covered (allowing for more pilot swaps) with the same number of photos or (B) using a number of handheld cameras, so all students may photograph. Students who are not collecting photographs should review their SfM Field Methods Manual and SfM Data Processing and Exploration Manual, work on assignments in their assignment packet (equipment list, workflow, worksheet, etc), collect metadata about the photo collection process (ex., sketch of target setup and planned camera locations/collection path, photo numbers) and typical field observations of the feature of interest. Plan a full afternoon for this portion to allow time for students to get individual time photographing. The processing time is directly dependent on the number of photos taken and the computer hardware used, so aim to collect roughly 100 to 150 photos to be able to generate a model overnight. However, it is useful to have extra photographs in case some are blurry. Have students measure three features on the outcrop (large clast, offset between stratigraphic units due to faulting, bed thickness, distance between two large trees are a few examples) that are clearly identifiable so they can later be compared to measurements made on the data set in the software.
*UAS regulations depend on location, as towns, cities, counties, and states have differing laws. Check local, state, and federal laws as well as university regulations for faculty UAS use prior to using a UAS to collect data for this module.
4) Creating the SfM Model
The field data collection portion of an SfM research project is a small component of the complete workflow. This step can be completed by individuals, pairs, or groups of students, as well as completed by an instructor with no student presence. As soon as field data collection has finished, use SfM software (the student exercise shows how to use Agisoft, but many other open-source options are available; see SfM Guide) to generate the 3D model of the field site. This could take eight to twelve hours, depending on the number and resolution of photos used as input. Make sure to georeference the model no matter the chosen workflow; this is necessary in order to measure features within the model. It is possible to load camera GPS data into Agisoft to speed up the model-generation process, specifically the first step of aligning photos. However, this data should be supplemented with ground control points from the GPS setup prior to surveying to achieve TLS-level accuracy.
5) SfM Data Exploration
This portion is optional but strongly encouraged, as students will need to be familiar with the software for subsequent units. If possible, the students should work with the data individually to maximize time with the software, but students may also work in pairs or in teams if computers are limited. One of the guides provided below should be distributed to support student work. Also attach a computer to a projector so students can follow the data exploration process as the instructor walks through the steps on their own machine.
Students should answer two reflective questions in their final write-up. One will cover the societal impetus/importance of using SfM to answer their research question of the day, and the other is a metacognition question so students can reflect on what they learned, an important aspect of the learning process. Another question you may have students answer is: What was rewarding and what was challenging about this exercise?
- SfM Research Applications Presentation (PowerPoint 2007 (.pptx) 67.3MB Oct31 16)
- Overview of ways that SfM is used
- Overview of ways that SfM is used
- SfM Workflow Presentation (PowerPoint 2007 (.pptx) 7.4MB Oct31 16)
- Basic information on how to design and conduct SfM surveys
- Basic information on how to design and conduct SfM surveys
- SfM Guide for Instructors and Investigators (Microsoft Word 2007 (.docx) 8.3MB Dec6 19) PDF (Acrobat (PDF) 9.1MB Dec6 19) - Guide to introduce instructors and researchers to Structure from Motion
- SfM Field Methods Manual for Students (Microsoft Word 2007 (.docx) 6.7MB Nov16 19) PDF (Acrobat (PDF) 8.7MB Nov16 19) - Includes SfM functions, considerations for survey design, and platform options
- Student Exercise
- Unit 1-SfM Student Exercise (Microsoft Word 2007 (.docx) 197kB May11 17) - Includes unit schedule, assignment sheet, blank equipment list, sketch page, notes page, and worksheet.
- Agisoft Metashape Pro - Creates point clouds from sets of overlapping photographs (Windows and Mac)
Agisoft software was selected because, at the time of writing, the majority of geoscientists were using it for SfM processing. Thirty-day fully-functional trial licenses have worked well for students to use on short-term assignments on their own computers. If you wish to use it repeatedly yourself or have it on laboratory computers, licenses will need to be purchased. Educational and multi-license pricing is available.
- Quick start guides - A short introductory guide is typically the best way for someone first using MetaShape
- MetaShape Quick Guide (Microsoft Word 2007 (.docx) 2.6MB Oct29 20) - This 4-page guide gets students started on a small SfM project in a local reference frame. It is from Getting started with Structure from Motion (SfM) photogrammetry) but can be modified for other types of initial exercises.
- MetaShape Guide for River Characteristics activity (Microsoft Word 2007 (.docx) 5.3MB Dec18 20) - This slightly longer quick guide was written specifically for Structure from Motion for Analysis of River Characteristics but is also a good entry point and can be modified for other types of exercises.
- Agisoft intro tutorial video - SfM - Agisoft tutorial (MP4 Video 21.8MB Mar8 17). Note: It was written for the earlier version of the software--PhotoScan. However the newer Agisoft software--MetaShape--is very similar so this video can still be useful.
- SfM Data Processing and Exploration Manual (Agisoft Metashape software) (Microsoft Word 2007 (.docx) 14.4MB Nov16 19)
- This more in-depth guide is better once one has some experience with the software
- Agisoft also provides many resources
- Quick start guides - A short introductory guide is typically the best way for someone first using MetaShape
- Other SfM processing software options are available and are outlined in SfM Guide for Instructors and Investigators (Microsoft Word 2007 (.docx) 8.3MB Dec6 19). Visual SfM written by another is below.
- Visual SfM Tutorial (Acrobat (PDF) 3.4MB Apr5 16)
- Visual SfM is an open-source software used for Structure from Motion. Our materials are based on using Agisoft MetaShape Pro, a commercial software. We provide this tutorial as an alternative to adapt for your students if MetaShape is not available for use.
- GETSI Field Prepared data sets- A number of SfM photos sets and contextualization information are available if field work is not possible. Other sources for high resolution topographic data sources include OpenTopography and UNAVCO Terrestrial Laser Scanning (TLS) Data Archive.
Teaching Notes and Tips
General advice on making the module work in field courses can be found on the module Overview page.
Remote field teaching
A simplified version of this unit is available that works for remote field situations. It can also be used in regular in-person teaching if the instructor wants a shorter exercise for introducing students to SfM - Getting started with Structure from Motion (SfM) photogrammetry (remote field collection).
In many cases it is very beneficial to use high-precision GPS to measure specific locations of ground control points. This definitely helps to generate a more accurate 3D model, comparable to the results produced by TLS. However, if GPS is not available, the GPS locations associated with each photo will work well enough to produce a less-accurate model (possibly with more internal distortion). In addition, you may use scale bars to integrate scale into your model. The value in investing effort in GPS alongside SfM depends on the geoscience application for which you are doing the survey and especially whether it depends on being tied to a global reference frame. Another module in the GETSI Field Collection covers field GPS - High Precision Positioning with Static and Kinematic GPS/GNSS.
Feature to survey
If teaching this unit/module in a classroom setting or on campus, try to choose an object/feature to survey that is similar to a geologic outcrop or feature. It is quite difficult to accurately survey statues, trees, or any other feature that requires 360° of data, and these should not be used as a feature in the introductory unit.
Although the structure of this unit differs from a usual day in a field geology program, remind students that the skills they learned in their previous weeks of the course (specifically, recording observations of the geologic feature of interest) should still be used in addition to the SfM survey. Metadata are an important aspect of survey design, as it is essential to keep good notes on the survey to ensure trouble-free processing and accurate interpretation of the data.
Keeping students occupied
One of the challenges of integrating SfM into a course with more than a few students is making sure that students stay engaged and mentally challenged even while they are waiting for their chance to participate in the data collection process. Students should also be encouraged to measure various clear features of the scanned area to ensure they can complete the data exploration portion of their assignment. Breaking students into small groups and conducting multiple surveys if multiples cameras and associated platforms are available is one good way to cut down on the amount of student free time.
A worksheet has been provided for students to calculate the area of their photos to determine the number of photos needed for a survey. The worksheet requires some knowledge about the camera used: the focal length, sensor dimensions (width and height), the aspect ratio of the photographs, and the effective megapixels of the camera. These variables may not be listed in the manual for the camera, but can be easily found using an Internet search.
If there is time prior to taking students in the field, have them find the feature of interest on Google Earth. This way, they can measure the scale of the feature and then use that measurement on the worksheet as well.
Adapting based on available computers
Part 5 of the unit (optional, but strongly encouraged) is designed for students to work on data visualization on individual machines. This section of the unit can be modified based on resources available. If no student computers are available, we encourage the instructor to project their screen and walk through the data exploration with students.
Adapting based on available software
The software used for Parts 4 and 5 of the unit is called Agisoft MetaShape Pro. This software is $549 per educational license (www.agisoft.com; price valid as of May 2021). The SfM Data Processing and Exploration Manual is written to guide students through using this software. However, there are many commercial and open-source alternatives. Of the commercial platforms, Agisoft is recommended for the ideal feature-to-price ratio. Open-source options include VisualSFM, a GUI similar to Agisoft, recommended for its ease of use and integration of many open-source SfM algorithms. A guide to VisualSFM written by Colorado State University graduate students is available as a PDF in the module materials.
Much of the formative assessment can be done through observations of and discussions with students individually, in pairs, or periodically in the whole group. Students can also hand in their work from the field, including field notes (atmospheric conditions, metadata) and sketch of survey setup. The work for formative assessment could be graded based on completion.
Summative assessment for Unit 1-SfM is based on the completed student exercise, which includes their workflow document, a summary of their metadata, details of the data collection, and results of initial data exploration. An assessment rubric is included in the student exercise. Summative assessment for the module as a whole will be evaluated at the end of the module in Unit 5. If the instructor is only planning to do Unit 1-TLS and then jump to Unit 5, the summative assessment we recommend that you choose from are "geologic outcrop with visible faulting," "channel sands," "dinosaur footprints," or other simple field site.
References and Resources
Consult these resources if additional information beyond the SfM Guide for Instructors (included in the unit materials) is needed.
General resources on SfM algorithms
- Lowe, D.G., (2004) Distinctive Image Features from Scale-Invariant Keypoints: International Journal of Computer Vision, 60 (2), 91-110, doi: 10.1023/B:VISI.0000029664.99615.94. This paper is the explanation of the algorithm that powers SfM software.
- Snavely, N., Seitz., S.M., and Szeliski, R., (2008) Modeling the World from Internet Photo Collections: International Journal of Computer Vision, 80 (2), 189-210, doi: 10.1007/s11263-007-0107-3. This paper was one of the first to implement the Lowe algorithm in easy-to-use software to generate large sparse point clouds of features like the Coliseum.
- Ullman, S., (1979) The Interpretation of Structure from Motion: Proceedings of the Royal Society of London, Biological Sciences, 203 (1153), 405-426, doi: 10.1098/rspb.1979.0006. The original paper on SfM.
General resources on SfM in practice
Many more resources than those listed below exist for examples of using SfM for a range of geoscience applications. A more exhaustive list is included in the SfM Guide for Instructors. However, new papers are appearing rapidly on the applicability of SfM, so a search is necessary to include all up-to-date research.
- Canon Hack Development Kit directions (Microsoft Word 2007 (.docx) 84kB Jun1 17) It is possible to load software onto camera memory cards so Canon Point&Shoot cameras (not DSLR) can be controlled in ways more helpful for doing SfM surveys. Canon Hack Development Kitis a unique software application developed by enthusiasts that enables extra features for ported Canon™ "Point&Shoot" cameras (not DSLR). This file provides a few extra details to clarify how to Canon Hack Development Kit.
- James, M.R., Chandler, J.H, Eltner, A., Fraser, C., Miller, P.E., Mills, J.P., Noble, T., Robson, S., Lane, S.N., (2019) Guidelines on the use of structure‐from‐motion photogrammetry in geomorphic research, Earth Surface Processes and Landforms, first published: 17 April 2019 https://doi.org/10.1002/esp.4637
- Bemis, S.P., Micklethwaite, S., Turner, D., James, M.R., Akciz, S., Thiele, S.T., and Bangash, H.A., (2014) Ground-based and UAV-based Photogrammetry: A Multi-scale, High-resolution Mapping Tool for Structural Geology and Paleioseismology, Journal of Structural Geology, 69, 163-178, doi: 10.1016/j.jsg.2014.10.007. Applications of SfM for structural geology as well as paleoseismology. This paper is a review paper and is a useful resource to demonstrate non-geomorphological applications of SfM.
- Dietrich, J.T., (2016) Riverscape Mapping with Helicopter-based Structure from Motion Photogrammetry, Geomorphology, 252, 144-157, doi: 10.1016/j.geomorph.2015.05.008. This researcher uses SfM frequently in their work. Go to his website (http://adv-geo-research.blogspot.com/)for useful tips on using SfM.
- Reitman, N.G., Bennett, S.E.K., Gold, R.D., Briggs, R.W., and DuRoss, C.B., (2015) High-resolution Trench Photomosaics from Image-based Modeling: Workflow and Error Analysis: Bulletin of the Seismological Society of America, 105 (5), 2354-2366, doi: 10.1785/0120150041. Reitman et al. show the importance of ground control points in the model generation process. This paper also outlines an SfM workflow, which may be a useful resource.
- James, M.R., and Robson, S., (2014) Mitigating Systematic Error in Topographic Models Derived from UAV and Ground-based Image Networks, Earth Surface Processes and Landforms, 39, 1413-1420, doi: 10.1002/esp.3609. One of the drawbacks of SfM is the lack of resources on testing model veracity. This paper shows the importance of camera location to model veracity.
- Johnson, K., Nissen, E., Saripalli, S., Arrowsmith, J.R., McGarey, P., Scharer, K., Williams, P., and Blisnik, K., (2014) Rapid Mapping of Ultrafine Fault Zone Topography with Structure from Motion: Geosphere, 10 (5), 969-986, doi:10.1130/GES01017.1. A comparison of differing software workflows for SfM as well as applications of SfM to active tectonics research.
- Additional recent published research: Reference list for research papers using SfM (Microsoft Word 2007 (.docx) 178kB Jul18 16)