In This Guide
Choose the payload class that fits the job
This table is the quickest place to start. It identifies the job each payload
class usually handles best and the mismatch most likely to waste money.
Scroll horizontally to compare all columns.
These categories overlap. A multi-sensor unit may combine wide, zoom,
rangefinding, and thermal modules, while an RGB camera may also support
photogrammetry. That does not remove the need to define the job. Every sensor
still has limits on distance, detail, calibration, mounting, and processing.
Define the result before comparing sensors
Write down what the customer, engineer, or asset owner must receive. "Collect
high-quality imagery" is too vague to guide a purchase. Better requirements
sound like these:
- Provide photographs in which a 5 mm crack can be reviewed from the permitted
flight distance.
- Deliver a georeferenced roof model checked against independent survey points.
- Record radiometric thermal files that a qualified analyst can use to compare
apparent temperatures.
- Classify vegetation using calibrated reflectance bands and field observations.
- Measure clearances or surface geometry in a registered point cloud.
The wording matters because different payloads measure different things. RGB and
zoom cameras record reflected visible light. Thermal cameras record infrared
radiation and, when radiometric, can estimate apparent temperature. Spectral
cameras separate reflected energy into selected wavelength bands. LiDAR
calculates range from emitted laser pulses. None of these outputs, by itself,
proves the cause of a defect.
It also helps to decide how far the result must go. Screening for an anomaly,
documenting visible condition, measuring geometry, and supporting an engineering
diagnosis are different jobs. The publication's guide to the
six levels of drone-inspection evidence
can help a team define that boundary before it buys hardware.
Use the right payload for each inspection scenario
General visual inspection
Start with an RGB camera when the work involves roofs, facades, towers, bridges,
equipment, or other assets that can be evaluated from color photographs or
video. It is usually the simplest option to fly, store, review, and share.
Choose the lens and flight distance around the smallest feature that matters. A
high-resolution wide-angle camera can still put too few pixels on a distant
fastener or crack. If the job also requires mapping, confirm that the aircraft
can collect the overlap, positioning information, and image consistency the
processing workflow needs.
Do not choose RGB alone when the required result is temperature, spectral
response, or direct range. Visible discoloration may support an inspection, but
it cannot substitute for the physical measurement the job calls for.
Detailed inspection from a safe standoff
Choose optical zoom when the aircraft must remain farther from energized
equipment, towers, bridge elements, or other difficult surfaces while still
showing small visual details. Optical zoom narrows the field of view and places
more detector pixels on the target. Digital zoom only enlarges pixels already
captured.
The tradeoff is stability and context. At long focal lengths, vibration,
pointing error, atmospheric effects, and focus become more visible. A narrow
view can also make it harder for the operator to maintain orientation. Confirm
that the gimbal remains stable at the required zoom setting and that the
workflow preserves both overview and detail images.
NIST's target-size guidance reinforces the practical point: recognizable detail
depends on how much of the frame the target occupies, not on the camera's
headline resolution alone.
Thermal inspection
Choose a radiometric thermal camera when the deliverable requires temperature
data for later analysis. A non-radiometric thermal view may help find patterns,
but it does not provide the same measurement record.
Detector resolution and lens choice determine how many thermal pixels cover the
target at the planned distance. Teledyne FLIR advises that a measurement target
should cover at least a 3 by 3 pixel area and overfill the measurement spot. A
component can therefore be visible in the image yet still be too small for a
reliable temperature reading.
Thermal measurements also depend on focus, emissivity, reflected apparent
temperature, atmosphere, viewing angle, calibration, and thermal contrast.
Choose this payload only when the team can control or document the variables
needed for the intended interpretation.
Vegetation or material analysis
Choose multispectral imaging when a known set of broad wavelength bands supports
a specific index, classification, or monitoring method. It is generally the more
manageable spectral option because the sensor, calibration process, storage, and
analysis are narrower in scope.
Choose hyperspectral imaging when fine spectral differences are central to the
job and the analysis team can support a much larger, more sensitive dataset.
NASA's remote-sensing training distinguishes spectral resolution, the number and
width of bands, from spatial resolution, the ground area represented by a pixel.
More bands do not ensure that a small target is spatially clear.
Both approaches require more than a camera. Plan for illumination changes,
radiometric calibration, band alignment, reference measurements, and field
observations. If the team does not have a validated model connecting spectral
data to the decision, collecting more bands may simply create more data to
manage.
3D geometry and low-texture surfaces
Choose LiDAR when direct range measurements, 3D structure, clearances, or
surface geometry are central to the deliverable, especially where image matching
may struggle with low texture or complex structure. LiDAR can also complement
imagery when the report needs both geometry and visual context.
Do not buy on maximum pulse rate alone. Point density, usable returns,
trajectory quality, scan angle, control, classification, and validation all
affect the finished point cloud. USGS quality levels specify pulse density and
vertical accuracy separately, which is a useful reminder that a dense point
cloud is not automatically an accurate one.
Ask the supplier to estimate usable point spacing on the actual surface. Narrow,
dark, wet, reflective, or steeply angled targets may produce less useful
coverage than a broad average points-per-square-meter figure suggests.
Check target size, distance, resolution, and accuracy
The smallest important feature should drive the sensor and lens calculation.
Record its real dimensions, the safe working distance, the viewing angle, and
how many pixels or points must cover it. Then ask the supplier or integrator to
show the calculation for the proposed configuration.
For frame cameras, ground sample distance is the physical distance represented
by one pixel at the target plane. It is useful for planning, but it does not
capture focus error, motion blur, compression, exposure, atmospheric effects,
oblique geometry, vibration, or lens performance. For a vertical asset,
calculate sampling at the asset surface rather than relying on a nadir mapping
value.
Resolution and accuracy are also different. Resolution describes how finely the
sensor samples the scene. Accuracy describes how close a measurement is to an
accepted reference. USGS guidance for UAS imagery notes that geometric accuracy
depends on ground control and tie-point quality regardless of image ground
sample distance. Mapping work still needs a coordinate reference, a control or
direct-georeferencing plan, independent checkpoints, and an agreed accuracy
statistic.
Apply the same distinction to the other payloads. Thermal resolution does not
guarantee temperature accuracy. Spectral resolution does not guarantee spatial
detail or correct classification. LiDAR point density does not guarantee
positional accuracy. Ask how the complete workflow will verify the result, not
merely how the sensor records data.
Confirm aircraft and gimbal compatibility
A payload purchase is also an aircraft-integration decision. Confirm each of
these items for the exact aircraft, payload, mount, and software combination:
- total mass, center of gravity, mounting interface, and gimbal clearance;
- power voltage, continuous draw, peak draw, startup behavior, and connector;
- command, trigger, time-synchronization, positioning, telemetry, and video
interfaces;
- onboard storage, write speed, file system, removable media, and transfer
method;
- supported flight-planning, camera-control, calibration, and processing
software;
- environmental limits, cooling, ingress protection, and electromagnetic
compatibility;
- expected endurance with the exact battery, payload, mounting, reserve policy,
and mission profile.
The FAA permits an external load under Part 107 only when it is securely
attached and does not adversely affect flight characteristics or
controllability. The aircraft, attached systems, payload, and cargo also remain
part of the applicable weight calculation. Meeting that rule is only the
starting point. The aircraft manufacturer or integrator must still support the
configuration and its operating limits.
Multi-sensor payloads can simplify mounting and control. DJI's H30 series, for
example, documents wide, zoom, rangefinding, and thermal modules in one payload
family. That example shows what an integrated architecture can include, but it
does not establish suitability for another aircraft, working distance,
environment, or deliverable.
Separate sensors can offer more flexibility but require closer interface work.
MicaSense's RedEdge-P integration guide documents power limits, connectors,
serial and Ethernet communication, triggering, and several integration modes.
Replace a supplier's broad claim of "compatibility" with a written interface
list. The
drone payload integration checklist
provides a more detailed review of the mechanical, electrical, timing, data,
environmental, and flight-test questions.
Payload mass and electrical load usually reduce flight time, but a generic
percentage is not useful. Request endurance for the exact configuration and
conditions you intend to fly. A payload that is technically mountable may still
be a poor purchase if it cuts each sortie below the time needed to collect the
job safely.
Budget for software, processing, and data handling
The sensor price is only one part of ownership. Estimate data produced per
flight, media write speed, transfer time, backup, processing hardware, software
licensing, calibration equipment, analyst time, training, and retention
requirements.
RGB inspection may need only selected stills and an asset-indexed report.
Photogrammetry adds overlap planning, reconstruction, coordinate systems,
control, checkpoints, and large derivative files. Thermal adds radiometric file
support and qualified interpretation. Multispectral work adds band registration,
reflectance calibration, and application models. Hyperspectral work adds heavier
spectral preprocessing and model management. LiDAR adds trajectory processing,
alignment checks, point-cloud classification, coordinate transformations, and
accuracy reporting.
Confirm who can open the native files and who will turn them into the final
deliverable. A payload that shortens collection but adds a specialist processing
queue may increase project time and cost.
Plan live communications separately from stored data. A compressed preview can
fit through a link even when the full-resolution dataset remains on the payload.
If the team needs real-time analysis, calculate latency and throughput across
the payload, aircraft, radio, ground system, and software. The guide to
drone bandwidth for video and payload data
explains why video-link bandwidth is not the same as the rate used to store or
transfer the complete dataset.
Test the complete system before purchase
Run a representative sample job with the proposed aircraft, payload, gimbal,
storage, software version, processing settings, and operator procedure. Include
the smallest target, hardest working distance, most difficult viewing angle, and
realistic environmental conditions. Use reference targets or independently
surveyed checkpoints when the deliverable involves measurement.
The test should answer four questions:
- Can the system show or measure the smallest target from the required distance
and angle?
- Does the saved dataset retain the necessary resolution, metadata,
calibration, timing, and position?
- Can the software produce the contracted deliverable within the available time
and computing budget?
- Does an independent check meet the accuracy or quality requirement agreed for
the job?
Treat a failed test as a purchasing disqualifier until the supplier demonstrates
a supported correction. Warning signs include unresolved motion blur, too few
thermal pixels on the component, missing radiometric data, sparse or uneven
point coverage, band misregistration, unsynchronized timestamps, unsupported
file formats, excessive power draw, unstable gimbal behavior, or processing that
misses the project schedule.
Questions to ask a payload supplier
Ask for answers tied to the exact model, lens, aircraft, mount, firmware, and
software versions you are considering:
- At the required distance, how many pixels or points will cover the smallest
target?
- Which measurements are radiometric, calibrated, or otherwise traceable, and
under what conditions?
- What aircraft and gimbal combinations are officially supported?
- What are the continuous and peak power requirements, including startup?
- Which trigger, timing, positioning, telemetry, video, and data interfaces are
supported?
- What flight time has been documented for this exact configuration and mission
profile?
- Which native files, metadata, calibration steps, and software licenses are
required?
- Can the supplier process a representative sample and deliver the files for
independent review?
- Which conditions, materials, angles, or temperatures cause degraded
performance?
- What training, maintenance, calibration, support, and replacement lead times
should be included in ownership cost?
A supplier that cannot answer these questions with model-specific documentation
is asking you to accept integration risk after the purchase.
Frequently asked questions
Can one drone payload handle both visual and thermal inspections?
Yes, some integrated payloads combine wide, zoom, and thermal cameras. That can
simplify mounting and keep visual and thermal context together. Confirm that
each module still meets the required target size, radiometric capability,
working distance, and file-output needs. A combined unit is convenient, but it
may not match the performance of a specialized sensor for every task.
Is a higher-resolution camera always better for inspection work?
No. The target must occupy enough pixels at the actual distance and viewing
angle, and those pixels must be sharp. Lens choice, focus, motion, exposure,
compression, atmosphere, and gimbal stability can matter as much as the total
pixel count.
When is LiDAR worth the added cost and processing?
LiDAR makes sense when direct 3D range, surface geometry, clearances, or
structure are central to the deliverable, or when image matching is unreliable.
If the client mainly needs clear visual documentation, RGB or zoom imagery may
be simpler and less expensive to collect and process.
How much flight time will a payload reduce?
There is no reliable universal percentage. The effect depends on aircraft,
batteries, payload mass, mounting drag, electrical draw, temperature, wind,
reserve policy, and flight profile. Require a documented estimate for the exact
configuration, then confirm it during a representative flight.
What should be verified before signing a purchase order?
Verify target coverage at the required distance, measurement limits, aircraft
and gimbal support, power and data interfaces, expected endurance, native file
access, software and processing requirements, environmental limits, calibration
needs, and the results of a representative sample job. Put any condition that
would make the payload unusable into the purchase criteria.
Choose the simplest payload that completes the job
Buy the least complex payload that can show or measure the target, operate
safely from the required distance, work with the aircraft, and produce files
your team can turn into the promised deliverable. If the team cannot define and
test that result before purchase, it is not ready to choose the sensor.
