Roadside Questions Answered

Automation can control repeatable capture, positioning or processing. A survey vehicle need not drive itself, and people remain responsible for validation and engineering decisions. Ask which tasks are automated rather than accepting “robotic” as a complete system description.

No. Cameras record visible surface features. Hidden structural weakness, obscured areas and some geometric properties require other evidence. Even visible distress can be missed when contrast is poor or a feature falls outside the system’s validated capability. Start with capture requirements.

Some systems use active lighting for low-light capture, but rain, reflections and standing water can obscure distress. Validate the operating conditions and repeat unusable sections. A sharp image of a wet surface is not necessarily a complete view of its condition.

No. Model scores depend on training and evaluation conditions. Check representative data, class-specific precision and recall, and human-reviewed samples. Include sections with no reported detections when reviewing missed findings. Our uncertainty guide explains the distinction.

Accurate enough for the intended task. Network screening and setting out a small repair need different tolerances. Record road, direction, lane and position quality. More decimal places do not establish better accuracy; see the mapping sequence.

No. A surface condition index summarises observations under a defined method. Structural capacity needs appropriate engineering assessment and potentially additional testing. A good-looking surface can conceal weakness, while visible distress does not by itself quantify remaining load-bearing capacity.

The interval depends on risk, deterioration, road function, available evidence and the operator’s inspection obligations. Routine network surveys do not replace urgent safety assessment. After severe weather, major works or a reported hazard, a targeted inspection may be necessary outside the normal cycle.

They can populate a workflow, but location, evidence quality, consequence and treatment suitability should be checked before work is authorised. Keep a record of who reviewed the finding and why the proposed action fits. The maintenance decision matrix provides a starting structure.

Include source imagery, stable identifiers, timestamps, road references, coordinate-system information, quality flags, review status and documented processing versions. Agree file formats, access permissions, retention and the handling of missing coverage before the survey begins.