What Should Be in a Drone Inspection Report?

On This Page

• The report contents to agree before flight
• Turn images into traceable findings
• Match the technical appendix to the claim
• Make coverage and uncertainty visible
• Specify a usable delivery package
• Questions to ask before accepting the report

The report contents to agree before flight

Use the following as an editorial checklist to adapt to the asset owner's inspection method. It is a proposed purchasing specification, not a universal industry standard. Minnesota DOT's 2018 research on improving bridge inspections with UAS explicitly focused on turning collected data into useful inspection deliverables for owners and engineers. The checklist applies that principle to the report handoff.

Before capture, the owner should supply the asset register or drawings, previous findings, component naming rules, required views, and the maintenance system's import fields. Agree who will interpret the data and who may assign condition or urgency ratings. Record missing inputs as limitations instead of quietly replacing the owner's asset IDs with a new numbering system.

This produces a concrete workflow: define the required observations, capture and check the images, preserve the originals, analyze the evidence, review the findings, then deliver and reconcile the files. Check focus, component identity and required views before leaving the site so the team can identify any recapture work while access remains available.

Turn images into traceable findings

A finding should lead a maintenance planner from a component to the supporting image without searching an entire flight folder. Give each finding a stable identifier and pair a context image with a detail image. The context establishes location; the detail shows the observation. Keep unmarked originals alongside annotated copies.

Use separate fields for observation, interpretation, and action. That separation lets another reviewer challenge a diagnosis without losing the underlying record. If automated software suggests a defect label, identify the suggestion and the human review status. An automated label should not silently become a confirmed condition.

A worked finding example

The following is a fictional reporting example, not a finding from an actual inspection:

• Finding: F-017, Building B, east roof edge, beside drain D-03.
• Evidence: context image B-E-014 and detail image B-E-015, both linked to retained originals and capture timestamps.
• Observation: a dark linear feature is visible beside the membrane seam in the detail image.
• Interpretation: cause undetermined from these images; the record does not establish an open seam or active leak.
• Measurement: none reported because the image set has no validated measurement basis for this feature.
• Action: facility maintenance lead to arrange a closer examination, set its due date under the owner's procedure, and record the result against F-017.
• Status: awaiting confirmation; urgency to be assigned under the owner's maintenance procedure.

The useful addition is the route to resolution. “Possible roof defect” alone leaves the recipient to locate it, decide what the label means, and invent the next step.

For repeat inspections, retain the finding ID and add dated observations. Record whether a change reflects a confirmed repair, a new view, a different analysis method, or a condition change. Do not treat a missing finding in the latest report as proof that the issue has disappeared.

Match the technical appendix to the claim

The appendix should contain the information needed to examine the actual conclusions. A visual observation does not automatically need a 3D model. A dimensional claim needs more than an annotated screenshot. Our guide to six levels of drone-inspection evidence explains the distinctions behind those choices.

Visual images and photogrammetric measurements

For visual findings, include sensor and lens identification, original image dimensions, relevant viewing conditions, and any processing that changes the presentation. State whether dimensions were measured or merely described. Avoid assigning a precise crack width from an image unless the capture geometry, scale and measurement method support it.

When a map or model supplies dimensions, ask for its coordinate reference system, units, software version, processing record, and supporting checks. For elevations, identify the vertical reference as well. A georeferenced result is not automatically accurate enough for the buyer's intended measurement.

For example, PIX4Dmapper's quality-report specification distinguishes ground sampling distance, ground-control errors, camera-position uncertainties, and reprojection error. Reprojection error describes how reconstructed points project back into images; it is not a direct measurement of a physical defect. Specify which reported statistic supports which conclusion, and request independent check measurements when the contract requires demonstrated accuracy.

Record a measurement's units, method, supporting check results and limitations beside the result or through a clear appendix reference. If uncertainty has not been evaluated, say so and leave its adequacy for the intended decision unresolved. A software-generated decimal does not justify the same precision in a maintenance recommendation.

Thermal findings

Thermal results need the matching visible image, the thermal original that retains measurement data, the temperature scale and units, and the parameters used for analysis. FLIR explains that calculated temperatures depend on emissivity, reflected temperature, ambient temperature, humidity and target distance; incorrect inputs produce incorrect calculations. See its drone thermography explanation.

Record those inputs and the conditions relevant to the inspection method, such as equipment operating state or recent weather. A colored thermal screenshot alone leaves the reviewer unable to check the calculation. Describe the observed thermal pattern separately from its proposed cause, and specify the examination needed to confirm an unverified diagnosis.

Do not apply one temperature-difference cutoff across unrelated assets. Ask the technical reviewer to identify the applicable method, comparison conditions and rating basis.

Make coverage and uncertainty visible

The report must distinguish examined with no reportable finding, finding recorded, and not adequately examined. Put coverage gaps in the summary and on the asset diagram, not only in a final disclaimer.

For example, an obstruction might prevent a view of the rear of a component even when its front is sharp. Mark the hidden surface as unexamined. Do not turn “no visible defect in the supplied views” into “component is sound.”

Technical quality can also be uneven within one dataset. Pix4D's quality-report guidance identifies blur, poor exposure and insufficient image overlap as potential causes of incomplete or poor reconstruction. Its guidance also distinguishes calibrated images from images not used in processing. For a model-based inspection, ask where image or reconstruction quality was insufficient and which findings depend on those areas.

Physical examination remains a separate issue. In its U.S. bridge inspection guidance, FHWA's answer to Q313-2 allows UAS to supplement inspection tasks but says they cannot address all aspects, including tactile examination, auditory cues and live-load response. An aerial bridge report should therefore state which tasks it supports and which required examinations remain with the inspection team. That bridge-specific guidance should not be presented as a universal rule for every commercial asset.

Keep the severity of an observed condition separate from confidence in the interpretation. A feature can need prompt review even when its cause is uncertain. Explain the owner's rating system and record who assigned the rating; a red marker has no useful meaning without that definition.

Specify a usable delivery package

Agree the files and access terms before quoting the work. The following are suggested contract choices, to be tested with the recipient's actual tools:

• Readable report: a searchable PDF with legible annotations, explanatory captions, revision identity and the agreed reviewer attribution.
• Findings register: a CSV or agreed spreadsheet with finding IDs, asset IDs, evidence filenames, status, action and responsible role.
• Image archive: originals with capture metadata retained, annotated derivatives, and a manifest connecting both to the register.
• Technical outputs when commissioned: thermal originals, a georeferenced orthomosaic, model or point cloud, accompanied by processing and measurement documentation.
• Access record: agreed download rights, hosting duration, retention period, permitted recipients and a procedure for obtaining the archive when platform access ends.

Treat format names as a starting point. Request a small representative export before award and have the receiving team open it. Check that a register row maps into the maintenance system, an image reference resolves after download, and a model uses the expected units and reference system. Do not assume compatibility because both parties say they support a geographic information system (GIS) or a 3D viewer.

Make delivery measurable with agreed dates for the initial report, technical review, corrections and final handoff. Count delivered assets against the scope, reconcile finding IDs across the report and register, and identify every exception. Image count and file size can document the delivery, but neither establishes inspection completeness.

Questions to ask before accepting the report

Ask the provider to demonstrate the handoff using an actual finding:

1. Can the recipient locate the component from the report and open its original evidence?
2. Does the coverage record distinguish missed areas from areas with no reportable finding?
3. Are measurements and ratings tied to a stated method and reviewer?
4. Are uncertain diagnoses and additional examinations clearly identified?
5. Do the downloaded files work in the receiving team's tools, with consistent IDs and units?
6. Can each unresolved item be assigned, updated and compared with the next visit?

Use the answers to accept the report, request a correction, or commission additional examination. A useful drone inspection report leaves the asset owner with located findings, understandable limitations, and work that can be assigned. Build those requirements into the purchase before the aircraft arrives.

Source notes

• Minnesota DOT, Improving the Quality of Bridge Inspections Using Unmanned Aircraft Systems (2018). Government-sponsored research record describing the focus on actionable inspection deliverables.
• Pix4D, Quality report specifications. Provider documentation defining processing metrics and report fields.
• FLIR, Four ways thermal reporting software supports drone inspections. Manufacturer explanation of thermal measurement inputs and retained radiometric data.
• Pix4D, Quality Report Help. Provider guidance on image quality, incomplete reconstruction and processing limitations.
• FHWA, National Bridge Inspection Standards questions and answers, Q313-2. U.S. government guidance on UAS as a supplement to bridge inspection.

Last checked: September 7, 2026.