Photogrammetry, Remote Sensing and Drone Surveying

Introduction

Surveying has traditionally obtained measurements by direct observation from the ground. Photogrammetry and remote sensing extend this capability by measuring objects and surfaces from photographs, multispectral images, radar, laser pulses, and other sensors. Unmanned aerial vehicles (UAVs), commonly called drones, have made these methods accessible for construction sites, quarries, roads, bridges, earthworks, and building documentation.

These technologies do not eliminate surveying principles. They depend on coordinate systems, control points, calibration, scale, adjustment, error analysis, and independent verification. A visually impressive orthophoto or three-dimensional model may still be geometrically unreliable if the camera was poorly calibrated, image overlap was inadequate, ground control was absent, or the processing settings were inappropriate.

In construction technology, remotely sensed data are used to document existing conditions, calculate stockpiles and cut-and-fill volumes, compare progress with design, inspect inaccessible assets, create terrain models, and maintain as-built records. Their responsible use requires technical competence as well as awareness of aviation law, privacy, occupational safety, data security, and professional liability.

Learning Objectives

After studying this chapter, the student should be able to:

1. Explain the geometry and terminology of close-range photogrammetry.
2. Distinguish ground control points, checkpoints, georeferencing, and camera calibration.
3. Explain orthophotos, digital elevation models, point clouds, meshes, and LiDAR.
4. Plan a basic UAV survey for construction or terrain documentation.
5. Estimate ground sampling distance and evaluate positional accuracy.
6. Identify sources of error and limitations in image-based surveying.
7. Assess regulatory, safety, ethical, and data-governance requirements.
8. Interpret remotely sensed outputs for progress monitoring, terrain analysis, and asset documentation.

1. Foundations of Photogrammetry and Remote Sensing

Photogrammetry is the science and technology of obtaining reliable geometric information about objects and environments from photographs. The essential principle is triangulation: a point appearing in two or more images can be located because the camera positions, viewing directions, and image coordinates constrain its possible three-dimensional position.

Remote sensing is broader. It refers to observing and measuring an object without physical contact. Passive sensors record naturally reflected or emitted energy, such as visible light, near-infrared radiation, or thermal radiation. Active sensors transmit energy and measure its return; LiDAR and radar are important examples.

Important terminology

• Image coordinate: The location of an image point, usually expressed in pixels or millimetres from the principal point.
• Object-space coordinate: The position of a point in the real-world reference system, such as eastings, northings, and elevation.
• Exterior orientation: The camera position and attitude when an image was captured.
• Interior orientation: The camera’s internal geometric parameters, including focal length and principal point.
• Tie point: A recognizable feature observed in two or more images.
• Ground control point (GCP): A marked or identifiable point whose coordinates are measured independently and used to georeference a project.
• Checkpoint: An independently surveyed point used to test accuracy, not normally used in adjustment.
• Point cloud: A set of three-dimensional points representing object surfaces.
• Orthophoto: An image corrected for perspective and terrain relief so that scale is approximately uniform and map measurements are possible.
• Digital surface model (DSM): A model of the visible surface, including buildings, vegetation, and other objects.
• Digital terrain model (DTM): A model intended to represent bare-earth terrain.
• Ground sampling distance (GSD): The approximate ground dimension represented by one image pixel.
• LiDAR: Light Detection and Ranging, an active laser-based measurement method.

2. Image Geometry

2.1 The central projection model

A conventional camera forms a perspective, or central, projection. Rays from object points pass through the perspective centre, commonly approximated by the camera’s optical centre, before reaching the image plane. Consequently, objects farther from the camera appear smaller, and parallel lines in the object may converge in the image.

For a level, flat surface and a simple camera model, the scale relationship is approximately:

where:

• GSD is ground sampling distance, in metres per pixel;
• H is camera height above the mapped surface, in metres;
• p is physical pixel size on the sensor, in metres per pixel; and
• f is focal length, in metres.

This relationship assumes a near-nadir image, a flat surface, negligible lens distortion, and a clearly defined height above the relevant ground. It is a planning approximation, not a complete accuracy equation.

If a sensor has a 4.8 μm pixel size, a focal length of 8.8 mm, and is flown 100 m above ground, then:

The expected GSD is therefore about 5.5 cm per pixel. A small GSD does not automatically guarantee 5.5 cm surveying accuracy; systematic errors, image matching, control, surface texture, and processing quality also matter.

2.2 Collinearity and bundle adjustment

The fundamental photogrammetric condition is collinearity: the image point, perspective centre, and corresponding object point lie on one straight line. In practice, a bundle adjustment simultaneously estimates camera orientations, camera calibration parameters, and three-dimensional point coordinates by minimizing image measurement residuals.

A simplified image-coordinate relationship is:

Here x,y are measured image coordinates; x₀,y₀ are principal-point coordinates; f is focal length; X,Y,Z are object coordinates; Xₛ,Yₛ,Zₛ are camera coordinates; and rᵢⱼ are elements of the rotation matrix describing camera attitude. The equations assume a pinhole camera model; real processing commonly adds radial and tangential lens-distortion terms.

The adjustment is usually overdetermined because many image observations contribute to fewer unknown parameters. Residuals should be inspected rather than hidden. A low image reprojection error can coexist with poor absolute accuracy when the model is internally consistent but incorrectly scaled or georeferenced.

Figure: Basic two-ray photogrammetric intersection.

2.3 Image overlap and network geometry

Forward overlap is the percentage of a flight line covered by successive images. Side overlap is the coverage shared by adjacent flight lines. High overlap improves the number of observations and helps maintain continuity, especially over irregular terrain. However, overlap alone cannot correct poor camera calibration, motion blur, featureless surfaces, or inadequate control.

A strong image network includes crossing flight lines, oblique images where vertical façades must be documented, and camera positions that provide suitable viewing angles. Nadir images are efficient for roofs and terrain; oblique images are essential for walls, bridge sides, stockpile edges, and building façades.

3. Ground Control, Georeferencing and Accuracy

A photogrammetric model initially has an arbitrary coordinate frame or may be weakly positioned by UAV satellite observations. Ground control connects the model to a known datum and scale. GCPs should be clearly visible, stable, evenly distributed, and measured with a suitable independent technique such as GNSS or total station surveying.

Control should not be concentrated only at the site boundary or in one corner. Points should cover the perimeter and interior and include elevation variation. Checkpoints must be withheld from the adjustment so that they provide an honest accuracy assessment.

Accuracy measures

For an error component such as east, north, or elevation, the root mean square error is:

where eᵢ is the error at checkpoint i, and n is the number of checkpoints. This assumes errors are being summarized consistently and that the checkpoints are representative. RMSE is not the same as a maximum error and should be reported with sample size, coordinate component, units, and test conditions.

A three-dimensional checkpoint error may be calculated as:

where eₓ, eᵧ, and e_z are coordinate differences in metres. Independent checkpoints are stronger evidence than software-reported control residuals.

Accuracy is affected by GSD, image geometry, camera calibration, control quality, satellite positioning, surface texture, lighting, motion blur, processing algorithms, and the chosen coordinate reference system. Survey specifications should therefore state both the intended product resolution and the tolerable positional error.

4. Outputs from Photogrammetric Processing

4.1 Orthophotos

A raw aerial photograph contains perspective displacement. A tall building leans outward from the image centre, and terrain elevation changes alter apparent position. Orthorectification uses camera orientation and an elevation model to project image information onto a selected reference surface. The resulting orthophoto has uniform map geometry within the limits of the model.

Orthophotos are useful for site layout, access planning, material mapping, environmental documentation, progress records, and comparing construction against design drawings. They may contain seams, colour differences, blurred areas, and occlusions. An orthophoto cannot show surfaces hidden from the cameras.

4.2 Point clouds and meshes

A dense point cloud is generated by matching common features across images. Each point may contain coordinates, colour, and sometimes confidence information. A mesh connects points into triangular surfaces and can support visualization, texture mapping, and model export.

Point clouds are particularly useful for measuring irregular earthworks and documenting complex structures. Their weaknesses include noise, holes, duplicated surfaces, vegetation contamination, and incorrect points generated from repetitive or reflective materials. Classification is required when separating ground, buildings, vegetation, and construction materials.

4.3 Elevation models and volumes

A DSM represents the uppermost observed surface. A DTM attempts to represent bare ground after filtering non-ground points. Confusing these products can cause serious volume errors. For example, a stockpile measured against a surface that includes vegetation or previous temporary material may produce a misleading result.

For a regular grid, an approximate volume between two surfaces can be calculated as:

where ∆zᵢ is the elevation difference at cell i, Aᵢ is the cell area, and V is volume. The method assumes the surfaces are correctly aligned and that the grid adequately represents terrain variation. Boundary definition and the choice of comparison surface often influence results more than the arithmetic.

5. LiDAR and Other Remote-Sensing Data

LiDAR emits laser pulses and measures the time or phase required for returned energy. For a time-of-flight system, range is related to:

where R is sensor-to-target range, c is the speed of light, and ∆t is the round-trip travel time. The factor 2 is required because the pulse travels to the target and back. The equation assumes the measured return is correctly identified and atmospheric and system effects are appropriately handled.

LiDAR can provide more reliable geometric information in low-texture environments than passive image matching. Some laser pulses penetrate gaps in vegetation, allowing ground classification beneath a canopy, although penetration is not guaranteed. UAV LiDAR systems are generally more expensive and require careful trajectory correction, calibration, and strip adjustment.

Multispectral sensors record several wavelength bands. Vegetation indices can assist environmental monitoring, while thermal cameras can support moisture or heat investigations. Such data require radiometric calibration and appropriate interpretation; colour differences alone should not be treated as proof of a material defect.

6. Planning a Basic UAV Survey

A sound survey begins with the decision, not the aircraft. Define the required deliverable, area, coordinate system, vertical datum, accuracy, date, access restrictions, and reporting format. Typical deliverables include an orthophoto, classified point cloud, DSM, DTM, contour map, cross-sections, volume report, or time-series comparison.

Planning workflow

1. Define purpose and specification. Establish the feature to be measured and allowable error.
2. Review the site. Identify boundaries, overhead cables, cranes, traffic, people, water, reflective surfaces, vegetation, and take-off or landing areas.
3. Select sensor and platform. A calibrated RGB camera may be adequate for general mapping; LiDAR, multispectral, or thermal sensors may be required for specialized tasks.
4. Choose altitude and GSD. Use the camera geometry approximation, then verify that the resulting coverage and accuracy meet the specification.
5. Design flight lines. Set suitable forward and side overlap; add cross-strips or oblique images for structures and vertical faces.
6. Establish control and checkpoints. Mark targets clearly and survey them in the project coordinate system.
7. Check weather and lighting. Avoid strong wind beyond aircraft limits, rain, fog, excessive shadows, and rapidly changing illumination.
8. Conduct a safety and regulatory review. Confirm pilot competency, permissions, airspace requirements, privacy obligations, and site controls.
9. Execute and record the flight. Log aircraft, sensor, time, altitude, speed, battery, weather, and anomalies.
10. Process and quality-control the data. Inspect image sharpness, overlap, control residuals, checkpoint errors, coverage, holes, and coordinate reference systems.
11. Archive and communicate limitations. Preserve raw data, processing settings, reports, and metadata.

Flight speed should be low enough to avoid unacceptable image smear. Exposure settings should generally prioritize sharpness, while automatic exposure changes should be controlled where radiometric consistency is important. A survey should be repeated when images are blurred, coverage is incomplete, or control targets are not visible.

7. Construction Applications

Progress monitoring

Repeated orthophotos and point clouds can show completed slabs, installed façade panels, excavation progress, temporary works, and material movement. Comparing dates requires consistent coordinate systems, camera planning, and surface definitions. A colour overlay may highlight differences, but engineering conclusions should be based on measured geometry and site records.

Terrain and earthworks

UAV data can produce contours, profiles, drainage studies, and cut-and-fill estimates. The survey team must define the perimeter, breaklines, stockpile bases, and treatment of vegetation. Volume comparisons are meaningful only when the dates and reference surfaces are compatible.

Asset documentation

Dense point clouds and textured models can record bridges, roofs, towers, façades, retaining walls, and construction details. Oblique imagery is especially important for vertical assets. Models should be labelled with date, datum, accuracy, and areas not observed.

Safety and inspection

UAVs can reduce the need for personnel to enter unstable slopes, high roofs, or active traffic zones. They do not remove risk: aircraft may collide with cranes or structures, batteries may fail, and inspection images may be misinterpreted. Defect diagnosis often requires close inspection, material testing, or specialist engineering judgment.

8. Illustrative Case: Earthwork Progress on a Highway Interchange

Background

A contractor is constructing a highway interchange involving large embankments and drainage channels. Monthly progress quantities are disputed because conventional cross-sections are sparse and stockpile boundaries change frequently. The project team proposes UAV photogrammetry.

Problem

The client requires a terrain model for progress measurement, with independent evidence that the results are fit for quantity assessment. The site includes active plant, haul roads, dust, steep embankments, and areas where vegetation has begun to grow.

Analysis

The team defines the required coordinate reference system, establishes twelve distributed GCPs, and reserves six additional checkpoints. A calibrated RGB camera is selected. Nadir imagery is planned for terrain, while oblique images are added around drainage structures and steep faces. The team identifies vegetation as a source of DTM error and plans ground classification and field verification.

Approach and solution

The pilot completes the regulatory and site safety checks, coordinates with plant operators, and conducts the flight during stable lighting. Image quality and coverage are checked before leaving the site. Processing produces an orthophoto, dense point cloud, DSM, classified DTM, and surface-to-surface volume comparison with the previous survey.

Checkpoint RMSE is calculated separately for horizontal and vertical components. Areas with vegetation, water, and active machinery are flagged. The quantity report states the survey date, boundary assumptions, coordinate system, control method, checkpoint results, software version, and excluded areas.

Outcome

The project obtains a repeatable record of embankment progress and identifies a drainage reach where construction differs from the design alignment. The measured volume is accepted for interim management after comparison with selected conventional checks. The report does not claim accuracy in obscured areas and recommends conventional surveying for final payment quantities where contract tolerances are stricter than the UAV results.

Learning points

The case demonstrates that the aircraft is only one part of the measurement system. Control, specification, independent checks, classification, metadata, and professional judgment determine whether the product is suitable for a contractual decision.

9. Limitations, Errors and Quality Assurance

Common failure modes include insufficient overlap, repetitive textures, transparent or reflective surfaces, moving machinery, water, dust, vegetation, shadows, poor satellite observations, weak control distribution, incorrect vertical datums, and inaccurate camera calibration. Image matching may fail on uniform concrete or freshly graded soil. Tall objects may be reconstructed with incomplete sides if only nadir images are used.

Quality assurance should include:

• checking image sharpness and exposure;
• confirming complete area coverage;
• verifying control and checkpoint coordinates;
• inspecting camera calibration and bundle-adjustment residuals;
• comparing independent checkpoints;
• checking for systematic tilt or doming;
• confirming units, coordinate reference system, and vertical datum;
• inspecting point-cloud classification;
• comparing selected measurements with total station, GNSS, or level observations; and
• documenting exclusions and uncertainty.

A model should never be treated as more accurate merely because it has millions of points. Point density is not equivalent to positional accuracy.

10. Regulation, Ethics and Professional Responsibility

UAV operations are governed by the aviation authority and local law. Requirements commonly concern aircraft registration, pilot competency, operating categories, maximum altitude, visual line of sight, operations near people or airports, night operations, airspace authorization, and flights over roads or property. The exact requirements vary by jurisdiction and must be checked before every project.

Construction sites also require risk assessment, exclusion zones, communication with workers, emergency procedures, and protection from rotor, battery, and collision hazards. A legal flight may still be unsafe if it interferes with cranes, traffic, or lifting operations.

Images can capture workers, neighbouring properties, vehicle registrations, security systems, and sensitive project information. Data collection should be necessary and proportionate. Access controls, retention periods, encryption, secure transfer, and controlled publication should be specified. Consent and privacy obligations may apply.

Professional responsibility includes honest reporting. Operators should not remove inconvenient points, alter measurements without traceability, or present a visualization as a certified survey. Deliverables should identify the person responsible, equipment, control, processing method, accuracy evidence, date, and limitations.

11. Contemporary Developments and AI

Automated image matching, semantic segmentation, object detection, and change detection increasingly support construction workflows. AI can classify roofs, vehicles, rebar, excavation areas, or safety conditions, but its outputs depend on training data, image quality, and site context. False positives and false negatives are possible, particularly when sites differ from the data used for training.

A responsible workflow treats AI as decision support. Critical dimensions, payment quantities, safety findings, and defect diagnoses require human review and independent evidence. Versioned models, audit trails, confidence information, and representative validation data improve accountability.

Integration with building information modelling (BIM), geographic information systems (GIS), digital twins, and project-control platforms can connect observed conditions with design objects and schedules. The benefits are greatest when coordinate systems, naming conventions, metadata, and data ownership are agreed before collection.

Chapter Summary

Photogrammetry determines three-dimensional information from overlapping images through central projection, triangulation, and bundle adjustment. Camera geometry provides useful planning relationships such as GSD, but resolution is not the same as accuracy. Ground control georeferences the model, while independent checkpoints provide a defensible accuracy test.

Orthophotos correct perspective and terrain effects for map-based interpretation. Point clouds, meshes, DSMs, DTMs, and LiDAR products support terrain measurement, volume calculation, progress monitoring, and asset documentation. UAV surveys require a defined specification, appropriate sensor, planned overlap, distributed control, safe flight execution, rigorous processing, and quality assurance.

Construction professionals must understand limitations caused by vegetation, water, texture, shadows, occlusion, calibration, and datum errors. Regulatory compliance, worker safety, privacy, data security, and honest reporting are essential. AI and digital-twin integration can improve efficiency, but professional judgment and independent verification remain necessary.

Key Terms

Photogrammetry: Measurement and interpretation of geometric information from photographs.

Remote sensing: Observation of objects without physical contact.

Central projection: Perspective image formation through a single projection centre.

Interior orientation: Internal camera geometry, including focal length and principal point.

Exterior orientation: Camera position and attitude at image capture.

Tie point: A common image feature observed in multiple photographs.

Ground control point: Independently surveyed point used to georeference a model.

Checkpoint: Independent point used to assess model accuracy.

Ground sampling distance: Ground dimension represented by one image pixel.

Bundle adjustment: Simultaneous estimation of camera and object parameters from image observations.

Orthophoto: Geometrically corrected image suitable for map measurement.

Point cloud: Collection of three-dimensional points representing surfaces.

Digital surface model: Elevation model including visible objects and terrain.

Digital terrain model: Elevation model representing intended bare ground.

LiDAR: Active laser-based ranging and mapping technology.

Georeferencing: Connecting data to a coordinate reference system.

RMSE: Root mean square error used to summarize measurement differences.

UAV: Unmanned aerial vehicle used for remotely piloted or autonomous flight.

Bundle geometry: Spatial arrangement of camera rays and image observations.

Change detection: Comparison of observations from different dates.

Review Questions

Short Answer Questions

1. What is the difference between a DSM and a DTM?
2. Why are checkpoints kept independent from ground control points?
3. Define ground sampling distance.
4. State two advantages of LiDAR over passive image matching.
5. Why are oblique images useful in building documentation?

Long Answer Questions

1. Explain the central projection model and describe how image overlap enables three-dimensional reconstruction.
2. Describe a complete workflow for planning and executing a UAV survey on a construction site.
3. Explain the role of control, calibration, bundle adjustment, and checkpoints in accuracy assurance.
4. Compare orthophotos, point clouds, meshes, DSMs, DTMs, and LiDAR products.
5. Discuss the safety, regulatory, privacy, and professional responsibilities associated with UAV surveying.

Analytical / Application Questions

1. A camera has a pixel size of 4.8 μm, focal length of 8.8 mm, and operates 100 m above level ground. Calculate the approximate GSD and explain why the result does not by itself establish survey accuracy.
2. A contractor reports a stockpile volume derived from a UAV point cloud. Identify at least six technical checks that should be completed before the quantity is accepted.
3. Design a survey strategy for a multi-storey building with a flat roof, complex façades, nearby traffic, and restricted airspace. Justify the imagery, control, safety, and quality-assurance decisions.

Multiple Choice Questions with Answers

Question 1

Which condition is fundamental to photogrammetric intersection?

A. The object point, camera centre, and image point are collinear

B. All images must be captured at night

C. The ground must be perfectly horizontal

D. The camera must use a thermal sensor

Correct Answer: A. The object point, camera centre, and image point are collinear

Question 2

Using GSD = (H × p) ÷ f, what is the approximate GSD for H = 100 m, p = 4.8 μm, and f = 8.8 mm?

A. 0.0055 m/pixel

B. 0.0545 m/pixel

C. 0.545 m/pixel

D. 5.45 m/pixel

Correct Answer: B. 0.0545 m/pixel

Question 3

What is the primary purpose of a checkpoint?

A. To increase the aircraft battery capacity

B. To colourize the point cloud

C. To independently test positional accuracy

D. To replace camera calibration

Correct Answer: C. To independently test positional accuracy

Question 4

Which product is intended to represent the visible upper surface, including buildings and vegetation?

A. Digital terrain model

B. Digital surface model

C. Ground control file

D. Orthographic datum

Correct Answer: B. Digital surface model

Question 5

Why are oblique images particularly valuable for a building survey?

A. They eliminate the need for any control points

B. They improve observation of vertical façades

C. They guarantee zero lens distortion

D. They prevent all shadows

Correct Answer: B. They improve observation of vertical façades

Question 6

In the LiDAR range relationship R = c × ∆t ÷ 2, why is the divisor 2 used?

A. The laser wavelength is doubled

B. The pulse travels to the target and back

C. Two cameras are always required

D. The aircraft flies at twice the planned speed

Correct Answer: B. The pulse travels to the target and back

Question 7

Which statement about point density is correct?

A. High point density automatically guarantees high accuracy

B. Point density has no relationship to data collection

C. High point density does not by itself prove positional accuracy

D. Point density replaces independent checkpoints

Correct Answer: C. High point density does not by itself prove positional accuracy

Question 8

Which practice gives the strongest evidence of absolute model accuracy?

A. Inspecting only a rendered 3D view

B. Using independent surveyed checkpoints

C. Increasing colour saturation

D. Deleting points with large residuals without review

Correct Answer: B. Using independent surveyed checkpoints

Question 9

What is a major limitation of a DTM derived from imagery over dense vegetation?

A. The sensor may fail to observe bare ground beneath vegetation

B. The coordinate system becomes automatically three-dimensional

C. The camera focal length becomes zero

D. The orthophoto cannot contain any pixels

Correct Answer: A. The sensor may fail to observe bare ground beneath vegetation

Question 10

Which is an appropriate professional practice for UAV deliverables?

A. Report the model without its datum or limitations

B. Present visualization as certified accuracy evidence

C. Document control, checkpoints, processing, accuracy, and exclusions

D. Remove all uncertain areas without recording the action

Correct Answer: C. Document control, checkpoints, processing, accuracy, and exclusions