Specify the observation first
Start with the smallest feature that matters to the intended decision, not with a camera’s headline pixel count. Network screening may tolerate a different level of detail from documenting narrow cracks before a sealing programme. State the required coverage, travel speed, working conditions and review process before choosing a sensor.
Ground sampling distance describes how much pavement is represented by an image pixel. It depends on the sensor, lens, mounting height and viewing angle. It is not a guarantee that a crack of the same width will be detected: contrast, blur, compression and the orientation of the feature also matter. Request examples that match the intended task.
Choose a capture arrangement
Area-scan cameras capture complete frames and can provide useful contextual views. Line-scan systems build an image from successive lines and require careful synchronisation with vehicle movement. Stereo arrangements use multiple views to estimate geometry, while laser-assisted profiling systems add a different measurement channel. None removes the need to check calibration and coverage.
| Arrangement | Useful contribution | Check before deployment |
|---|---|---|
| Area-scan camera | Context and repeated surface frames | Overlap, shutter behaviour and motion blur |
| Line-scan camera | Continuous high-detail strips | Distance triggering and illumination uniformity |
| Stereo cameras | Geometry from paired views | Calibration stability and weak-texture areas |
| Laser-assisted imaging | Surface profile or texture information | System specification, validation and safe operation |
Compare the delivered evidence, not just sensor specifications. A system with more pixels can still produce less useful imagery if its mounting arrangement admits vibration, its exposure is too long or its processing discards fine detail.
Control light and motion together
Short exposure times can reduce motion blur but demand adequate light. Active illumination can improve consistency, although it must be designed for the sensor, working distance and operating environment. Strong sunlight, deep shadows and wet reflective surfaces remain important test conditions. Record them rather than assuming one configuration will handle every route.
A global shutter exposes the image together; a rolling shutter reads different rows at different times. Movement can therefore affect geometry differently. The practical question is whether the complete capture system produces stable, interpretable images at the planned operating speed. Test mounting vibration, focus and exposure across that speed range.
Do not treat “all-weather” as a universal capability. A road film of water may conceal texture even when the camera still returns a sharp picture. Establish stop, repeat-survey and review rules for rain, standing water, debris and low visibility. Safety and local road-working requirements remain separate obligations.
Keep calibration traceable
Document camera identity, lens settings, mounting dimensions and calibration dates. Intrinsic calibration concerns the camera’s internal geometry, including lens distortion. Extrinsic calibration describes its position and orientation relative to other sensors or the vehicle. Changes to mounts or lenses can invalidate assumptions used downstream.
Synchronisation also matters. A clear frame with the wrong timestamp may be assigned to the wrong road location. Check the relationship between camera triggers, positioning records and distance measurements. Keep original timestamps and record any processing offsets so that a later reviewer can reconstruct the chain.
Accept a survey through evidence
Agree a field acceptance procedure before production capture. Include representative asphalt, concrete, repaired areas, markings and difficult lighting. Review original images as well as processed outputs. A route-level coverage summary should identify missing frames, unusable strips and sections needing another pass.
Use a repeatable control section to compare runs, but do not mistake repeatability for accuracy. Independently inspect a representative sample and retain the disagreements. ISO 55000 provides an asset-management context for specifying information needs; ISO 19157-1 provides a framework for describing geographic data quality. Neither is a substitute for a project-specific camera acceptance test.
Once the images are fit for purpose, move to defect interpretation. The strongest capture specification is one that makes later uncertainty visible rather than hiding it behind a polished image.