🚧 Under construction — VoTechWorks is being built. A free tech program — complimentary, as our commitment to get America working.
voTechWorksGrade 13 · a vocational technology learning center, by urbanicity
newSchoolWorks.com › VoTechWorks › Technology & Design › GIS / Surveying / Drone Operations › Photogrammetry: Aerial and Close-Range Measurement › Lesson 1
TEC-GIS45-06 · Lesson 1 · 10 clock hours · theory

Photogrammetry Principles and the SfM Workflow

Objective. Students explain the geometry of stereo overlap, the structure-from-motion algorithm, and the sequence of steps from image capture to final deliverable.

⚠ Pending professional review

This lesson, its worked examples and its practice questions has not yet been verified by an industry-certified professional or by a certification or approval body. It is a study draft and may contain errors. Do not use it as instructions for work on a job site or on live equipment. Always follow your instructor, your supervising professional, and current official references and regulations.

VoTechWorks program approval and accreditation are still in progress. Credentials are issued only by the certifying bodies named in each program, after their own exams.

What Is Photogrammetry?

Photogrammetry is the science of making measurements from photographs. By analyzing the geometry of overlapping images, we can reconstruct the three-dimensional shape, size, and position of objects and terrain. In GIS and surveying work, photogrammetry converts drone or aircraft imagery into deliverables such as digital surface models (DSMs), digital terrain models (DTMs), and orthomosaics that can be brought directly into ArcGIS Pro or other GIS platforms.

Stereo Overlap Geometry

The foundation of photogrammetry is stereo overlap — capturing the same ground feature from at least two camera positions. When two images share a common area, corresponding points in each image can be matched and their three-dimensional position calculated through triangulation.

Forward overlap (also called end lap) is the percentage of one image that is duplicated in the next image along the flight line. Standard practice for drone mapping is 75–80% forward overlap. Sidelap is the percentage of overlap between adjacent flight lines; 60–70% is typical. Higher overlap improves point-cloud density and reduces data gaps but increases flight time and processing load.

The base-to-height (B/H) ratio describes the relationship between the distance separating two camera stations (the base) and the flying height above the subject. A larger B/H ratio improves depth resolution — the ability to measure elevation differences precisely — but also increases geometric distortion and the risk of occlusion (hidden areas). For most mapping missions, a B/H ratio between 0.3 and 0.6 balances accuracy with practical coverage.

Convergent imaging means cameras are angled toward the subject from multiple directions rather than all pointing straight down (nadir). Adding oblique or convergent passes helps reconstruct vertical surfaces, building facades, and complex terrain that nadir-only flights miss.

Structure-from-Motion (SfM): How It Works

Structure-from-Motion is a computational photogrammetry technique that simultaneously estimates camera positions and a sparse 3-D point cloud from a set of overlapping images — without requiring a fixed, known camera geometry in advance. The name comes from the fact that 3-D structure is derived from the apparent motion of the camera through a scene.

SfM relies on feature-detection algorithms (commonly SIFT or similar) to find distinctive points — tie points — that appear in multiple images. By matching these points across images, the software solves for the position and orientation (exterior orientation) of every camera at the moment of exposure, as well as any lens distortion parameters (interior orientation / camera calibration). This joint solution is called bundle adjustment.

The SfM Workflow: Step by Step

1. Image Acquisition. Plan and fly the mission with sufficient overlap, appropriate ground sampling distance (GSD), and consistent lighting. Collect ground control points (GCPs) with a GNSS receiver or total station if the project requires survey-grade accuracy. Safety note: follow all FAA Part 107 rules — check airspace authorization, maintain visual line of sight, and conduct a pre-flight safety check of the drone and batteries before every flight.

2. Tie-Point Extraction. The software scans every image for distinctive features and matches corresponding features across overlapping images. The result is a sparse set of matched points that link the image network together.

3. Bundle Adjustment. A least-squares optimization simultaneously refines camera positions, orientations, and lens parameters so that the reprojected tie points minimize error across the entire block of images. If GCPs are provided, they anchor the model to a real-world coordinate system. The reprojection error (measured in pixels) is a key quality indicator; values well below 1 pixel are desirable.

4. Dense Matching. With camera geometry solved, the software computes depth for every pixel, producing a dense point cloud containing millions or billions of points. This step is computationally intensive and requires a capable workstation with a modern GPU.

5. DSM and Mesh Generation. The dense point cloud is interpolated into a continuous digital surface model. A textured 3-D mesh may also be built for visualization or engineering use.

6. Orthomosaic Export. The software orthorectifies each image — removing terrain and camera tilt distortions — and mosaics them into a single, geometrically correct raster image. The orthomosaic has a consistent scale and can be measured like a map.

7. GIS Integration. The DSM and orthomosaic are imported into ArcGIS Pro (or another GIS) for analysis, editing, and sharing. Coordinate systems must match the project datum; mismatched projections are a common source of error at this stage.

Accuracy Terminology and Standards

The American Society for Photogrammetry and Remote Sensing (ASPRS) Positional Accuracy Standards for Digital Geospatial Data define how accuracy is reported and tested for geospatial products. Key metrics include Root Mean Square Error (RMSE) computed from check points — independent points not used in the bundle adjustment. Horizontal and vertical accuracy are reported separately. Understanding these standards is essential for communicating deliverable quality to clients and for meeting contract specifications.

Ground sampling distance (GSD) — the real-world size of one image pixel on the ground — directly limits the spatial detail and accuracy achievable. A finer GSD (smaller number) requires flying lower or using a higher-resolution camera.

Safety and Data Integrity Notes

Always complete a thorough pre-flight inspection checklist before drone operations. Keep firmware and software updated, and verify battery health. On the ground, protect collected imagery and GCP data with regular backups before processing begins — raw data loss after a flight is irreversible. During office processing, label all project files clearly with date, location, and coordinate system to prevent costly errors in GIS integration.

Key terms10

Forward overlap (end lap)The percentage of an aerial image that is repeated in the next image along the flight line; typically 75–80% for drone mapping to ensure sufficient tie points.
SidelapThe percentage of overlap between images from adjacent flight lines; typically 60–70% to prevent gaps between strips.
Base-to-height (B/H) ratioThe ratio of the distance between two camera stations (the base) to the flying height above the subject; influences depth measurement precision.
Structure-from-Motion (SfM)A photogrammetric technique that simultaneously recovers camera positions and a sparse 3-D point cloud from overlapping images by tracking how features move across frames.
Bundle adjustmentA least-squares optimization that simultaneously refines all camera positions, orientations, and lens parameters so that reprojected tie points best fit across the entire image block.
Ground control point (GCP)A point on the ground with a precisely measured real-world coordinate, used to anchor and scale the photogrammetric model to a known coordinate system.
Ground sampling distance (GSD)The real-world size represented by one pixel in an aerial image; a smaller GSD means finer spatial detail and is achieved by flying lower or using a higher-resolution sensor.
OrthomosaicA geometrically corrected, mosaicked aerial image from which accurate measurements can be made, produced by orthorectifying individual images and stitching them together.
Reprojection errorThe distance, measured in pixels, between where a tie point was observed in an image and where it is predicted to appear based on the solved camera geometry; a key quality indicator in bundle adjustment.
Dense matchingThe SfM stage in which depth is computed for every pixel using the solved camera geometry, producing a dense point cloud that represents the surface in detail.

Worked examples4

1. A drone flies at 120 m above ground level with a camera that has a sensor width of 17.3 mm and a focal length of 24 mm. What is the approximate ground sampling distance (GSD) in centimeters?
  1. Identify the formula: GSD = (sensor pixel pitch × flight height) / focal length. First find pixel pitch. The sensor has 5472 pixels across 17.3 mm, so pixel pitch = 17.3 mm ÷ 5472 ≈ 0.00316 mm = 0.00000316 m.
  2. Convert flying height to the same units: 120 m.
  3. GSD = (0.00000316 m × 120 m) / 0.024 m = 0.0003792 m / 0.024 = 0.01580 m.
  4. Convert to centimeters: 0.01580 m × 100 = 1.58 cm.
Answer: The GSD is approximately 1.58 cm per pixel, meaning each pixel represents about 1.58 cm on the ground.
2. A mapping mission is planned with 80% forward overlap and 65% sidelap. The camera footprint on the ground is 200 m (along-track) × 150 m (across-track). What are the along-track photo spacing and the across-track strip spacing?
  1. Along-track spacing = footprint length × (1 − forward overlap) = 200 m × (1 − 0.80) = 200 m × 0.20 = 40 m.
  2. Across-track strip spacing = footprint width × (1 − sidelap) = 150 m × (1 − 0.65) = 150 m × 0.35 = 52.5 m.
Answer: Photos should be taken every 40 m along the flight line, and flight strips should be spaced 52.5 m apart.
3. Two camera stations are separated by a base distance of 60 m, and the flying height above the terrain is 150 m. What is the base-to-height (B/H) ratio, and is it within the commonly recommended range of 0.3–0.6?
  1. B/H ratio = base / height = 60 m / 150 m = 0.40.
  2. Compare to the recommended range: 0.40 is between 0.3 and 0.6.
Answer: The B/H ratio is 0.40, which falls within the recommended range of 0.3–0.6, indicating a good balance between depth accuracy and coverage.
4. After bundle adjustment, a project report shows a reprojection error of 2.8 pixels. The project manager says this is too high and asks what it likely indicates. Describe two probable causes.
  1. A reprojection error of 2.8 pixels is elevated (values well below 1 pixel are desirable). Possible cause 1: insufficient or poor-quality tie points due to low texture in the imagery (e.g., uniform vegetation or water surfaces), causing the feature-matching algorithm to produce incorrect correspondences.
  2. Possible cause 2: inaccurate or poorly distributed ground control points (GCPs), or GCP coordinates entered with errors, causing the bundle adjustment to fail to converge on a consistent geometric solution.
Answer: A reprojection error of 2.8 pixels likely indicates either poor tie-point matching (low image texture or motion blur) or problems with GCP quality, quantity, or distribution. Both issues would cause the bundle adjustment to produce a poorly fitted camera network.

The math in this lesson2

Learners compute pixel pitch and GSD = pitch x height / focal length, about 1.58 cm at 120 m.
A-SSE.1N-Q.1
Brush up in the math halls at newSchoolWorks.com:
In class: Lecture and annotated diagram exercise tracing a ray bundle through camera stations to a reconstructed 3-D point, labeling base-to-height ratio, overlap, sidelap, and convergence. · Small-group reading and discussion of an ASPRS accuracy standards summary and an Esri ArcGIS Pro photogrammetry workflow guide, then each group presents one stage of the SfM pipeline to the class. · Written concept map: students draw the full workflow (image acquisition → tie-point extraction → bundle adjustment → dense matching → DSM/orthomosaic export → GIS integration) and annotate each step with its purpose and a possible error source.
Materials: Whiteboard, ASPRS Positional Accuracy Standards for Digital Geospatial Data (publicly available PDF), Esri ArcGIS Pro documentation, Agisoft Metashape documentation, projector
Check: Exit quiz of 10 multiple-choice questions covering overlap geometry, SfM stages, and accuracy terminology; pass = 70%.

Practice check8

Answer each question, then check it. Nothing is recorded; this is your own practice.

Photogrammetry / Esri ArcGIS Pro
1. Which term describes the percentage of an image that is repeated in the next image taken along the same flight line?
Photogrammetry / Esri ArcGIS Pro
2. In the SfM workflow, what is the primary purpose of bundle adjustment?
Photogrammetry / FAA Part 107
3. A drone is flown at 100 m AGL with a camera sensor pixel pitch of 0.004 mm and a focal length of 20 mm. What is the approximate GSD?
Photogrammetry / Esri ArcGIS Pro
4. Which of the following best describes what an orthomosaic is?
Photogrammetry / NSPS CST L1-2
5. A project has a B/H ratio of 0.15. Compared to a ratio of 0.45, what is the most likely consequence?
Photogrammetry / NSPS CST L1-2
6. According to standard photogrammetric practice, which independent measurements are used to compute RMSE for reporting positional accuracy of a geospatial product?
Photogrammetry / Esri ArcGIS Pro
7. Which stage of the SfM pipeline directly precedes dense matching?
FAA Part 107
8. Before launching a drone mapping mission under FAA Part 107, a remote pilot must:
Record this lesson on my transcript. Sign in with your key and enrollment email to record lessons (enroll free if you have no key). This page stays open to read and practise.
Program planNext lesson →