On This Page
What is actually automated?
Separate three functions when evaluating an inspection system:
- Flight: following a route or selecting a path around the asset.
- Capture: choosing viewpoints, camera orientation, spacing, and when to
take photographs.
- Interpretation: identifying possible defects, measuring features, or
assigning findings for review.
A waypoint mission can automate flight while leaving the pilot to operate the
camera. An adaptive scanning system can plan viewpoints from sensed geometry.
Neither description establishes that the system can diagnose the asset's
condition. Manual piloting here means a person directing the aircraft and
viewpoints, potentially with stabilization and obstacle assistance still active.
Skydio's
X10 3D Capture instructions
provide a concrete example: the operator defines a scan volume, and the software
develops flight patterns within the capture workflow. The operator can pause and
take additional photographs; manual photos join the scan's image set. That
documented combination illustrates why automated and manually directed capture
can belong in the same inspection.
Ask a supplier to name each automated function, its required inputs, and what
the operator must still decide. Our
six-part test for automation and authority
helps distinguish execution of a task from authority to handle exceptions.
How the workflows compare
Use the same asset, inspection objective, camera capability, required detail,
and reporting scope when comparing methods. The table describes workflow
differences and their practical implications. It is an editorial comparison
informed by the Skydio capture instructions and
PIX4Dmapper's image-acquisition guidance,
checked September 7, 2026; it is not a measured performance ranking.
Scroll horizontally to compare all columns.
The work shifts rather than disappears. Automation places more of the collection
logic into mission preparation and configuration. Manual capture places more of
it into decisions during flight. Both need a way to reconcile what was requested
with what was delivered.
A field preview can help with that reconciliation. Skydio documents an
Onboard Modeling preview
built from image thumbnails, plus a coverage heatmap. Use such a preview to
identify areas needing another look. Open the original photographs before
accepting fine detail; a quick model preview serves a different purpose from a
finished inspection deliverable.
Accuracy: repeatability is only one part
An automated route can help repeat a capture pattern. Whether the resulting
inspection is accurate depends on the question being asked. Separate visible
detail, measurement error, and correctness of interpretation.
Can the reviewer see the required feature?
For a photographic inspection, specify the component face and the detail that
must be recognizable. Review original images for focus, obstruction, glare, and
blur. A completed route does not prove that a recessed connection or the back of
a member was visible.
For photogrammetry, which reconstructs geometry from overlapping photographs,
capture geometry matters. PIX4D's guidance explains that the image plan must
suit the object and that building reconstruction needs views around the
structure. A downward-looking site survey is therefore a different assignment
from gathering detailed facade views. Manual or automated flight can produce
suitable images if the capture plan meets that need.
Can the measurement be trusted?
Ground sample distance is the ground distance represented by one pixel, commonly
expressed in centimeters per pixel; it is not a measurement-error guarantee. The
USGS imagery-calibration report
explains that control-point accuracy and tie-point quality affect geometric
accuracy regardless of ground sample distance. Tie points are matching features
used to connect photographs during processing.
For dimensional work, require checks against independent reference measurements
appropriate to the project. Document the reference system and the uncertainty of
the checks. Repeat the same checks for both methods. A repeatable image set can
still contain a repeatable systematic error.
Thermal inspection adds another measurement chain.
FLIR's thermographic guidance
identifies emissivity, reflected radiation, distance, and atmospheric conditions
as inputs affecting temperature measurement. Emissivity describes how
effectively a surface emits infrared radiation. Repeating a camera position does
not establish that those inputs or the asset's operating condition stayed
comparable.
Does the finding mean what the report says?
Keep flight-mode performance separate from defect-analysis performance. If
software labels defects, ask for a separate evaluation against examples reviewed
by a qualified asset specialist. Record both missed findings and false alarms,
together with conditions where an image could not be assessed. Do not let a
single overall percentage hide missed consequential defects.
Neither manually directed nor automated visual capture substitutes for every
inspection technique. In its U.S. bridge guidance,
FHWA describes UAS as supplementary
and identifies work such as tactile examination and sounding that they cannot
perform. Changing the flight mode does not close that evidence gap.
Limitations that change the choice
Incomplete or outdated geometry: a reusable mission needs review when
scaffolding, equipment, vegetation, or the asset itself changes. Check whether
the system merely replays coordinates or senses the current scene and adapts.
Plan how a blocked view will be recorded and revisited.
Navigation and obstacle-sensing limits: obtain the limitations for the
actual aircraft, flight mode, and environment. Skydio's X10 capture
instructions, for example, state that obstacle avoidance does not detect moving
objects and is disabled during landing. Do not generalize that behavior to every
platform, or assume manual control makes an unsuitable environment acceptable.
Missing recovery options: establish who can pause the mission, take control,
or terminate it, and what happens after communications loss. Include the crew
needed for those responsibilities in the operating model. A dock does not, by
itself, establish a workable unattended inspection service.
Operating permission: for a U.S. Part 107 operation, account for
visual-line-of-sight requirements and any necessary waiver or airspace
authorization. The
FAA's commercial-operator guidance
identifies these requirements and waiver paths. Automatic flight capability does
not supply the permissions for a proposed remote operation.
Data handoff: request a sample export before committing to a workflow. It
should preserve original evidence, capture time, asset and component identity,
finding location, and review status. Test whether the receiving maintenance team
can open the files and connect a finding to its work-order process. Treat extra
manual labeling and failed imports as delivery work in the cost comparison.
Compare costs per accepted inspection
Compare the cost of completing the same inspection scope, including corrections
and review. Use a common period and currency, and separate one-time setup from
recurring work. An automated method may justify more setup when a mission is
repeated; an isolated inspection may never use that investment again.
Total period cost = allocated aircraft and sensor cost + mission setup +
software and integration + field labor and travel + processing and technical
review + maintenance and support + correction visits.
Cost per accepted inspection = total period cost / inspections accepted against
the agreed scope.
These are costing formulas, not market-price estimates. Populate them with your
quotes and recorded labor. Define what counts as an accepted inspection before
comparing totals; otherwise, a cheaper partial survey can appear to beat a
complete one.
For automation, ask about mission configuration, subscriptions, supported
exports, training, and any dock installation or connectivity needed. For manual
work, account for pilot and inspection-specialist time, travel, and image
organization. In both cases, include processing and review: collecting
additional photographs can create work after landing.
For a recurring program, a simple break-even calculation can clarify the
purchase discussion:
Break-even inspection count = additional automation setup cost / (manual
recurring cost per inspection - automated recurring cost per inspection).
Use that expression when automation has a positive additional setup cost but a
lower recurring cost, both methods deliver the same accepted scope, and the
recurring figures include the same cost categories. Round up to a whole
inspection. If the denominator is zero or negative, recurring savings cannot
recover the added setup cost. Rework, mission changes, financing, and
subscription commitments may require a fuller model.
The largest uncertainty may be exception work. Ask how many visits in a proposed
program still require local attendance, targeted recapture, or another
inspection method. Do not subtract those costs merely because the main flight
sequence is automated.
Best-fit commercial missions
These are selection recommendations derived from the capture and evidence limits
above, not guarantees for a particular aircraft or site.
Scroll horizontally to compare all columns.
For example, a roof program might use planned capture for each scheduled
baseline and manually directed views for a suspicious flashing detail. Keep the
added photographs tied to the same roof section and visit. Otherwise, the
flexibility that helps investigate the finding can make later comparison harder.
Run a matched trial before scaling
Choose a representative asset section containing both straightforward coverage
and difficult views. Give each method the same component list, required detail,
and delivery specification. Keep the sensor configuration, capture conditions,
and processing approach comparable, or explicitly record why they differ.
Have the technical reviewer assess the delivered material, then compare:
- Required views accepted, missing, or needing correction.
- Independent measurement checks, if dimensional results are required.
- Follow-up needed to resolve observations or ambiguous images.
- Total person-hours from preparation through accepted handover.
- Export usability and the ability to recreate the inspection on a later visit.
Log interventions and corrections rather than counting only uninterrupted
flights. Test a repeat visit if repeatability is central to the business case.
Choose automation where the trial shows that repeatable capture reduces total
work while preserving the required evidence. Choose manual piloting where the
task is selective and the useful viewpoint develops during inspection. Keep a
hybrid workflow when systematic coverage and targeted investigation both
contribute to the maintenance decision.
Source notes
Last checked: September 7, 2026.
