Zum Inhalt springen Skip to content

Room book

Confidence and check log in automated area measurement

Updated Deutsch

A check log documents, for every automatically measured area, where it comes from and how certain the detection was, so that a third party can verify it without measuring again. Confidence here is a measure of detection certainty – not of accuracy against the building. Keeping the two apart is the core of a traceable process.

In short

Definition

A check log for areas taken from plans contains at least: the plan source with revision and origin of the scale (dimension chain, stated scale, estimate); the detected geometry per room with source reference in the original plan; the standard and calculation rule applied (DIN 277, WoFlV); the confidence per item; flags for uncertain detections, incomplete dimension chains and rooms without a closed outline; editor and sign-off with date.

Confidence arises in the detection step – a model rates how unambiguous a wall, a door, a room was in the image. Accuracy is a property of the result against reality and can only be established by comparison with a measurement. A plan can be read with high confidence and still be wrong if the building was built differently.

Distinction

Term What Who measures it What for
Confidence certainty of detection in the plan the software, per item order of checking: uncertain items first
Accuracy deviation of the result from the building comparison with the survey statement whether the figure holds
Tolerance permissible deviation under a set of rules or a contract standard, contract, case law decision whether a deviation has consequences

In practice

Anyone reading a check log starts with the low confidences and the flags – that is where it is decided whether the result holds. For documented proof under VOB, a living area calculation or the initial population of a CAFM system, the log is the answer to the question “Where does this figure come from?” – and the evidence that detection and calculation are separate: detection reads the plan, a calculation computes deterministically, and every step is documented. Pure language models do not deliver this; they estimate instead of calculating.

Normraum relevance

In Normraum, every room carries a confidence and a source reference in the original plan; the check log documents standard and calculation path and flags uncertain detections and incomplete dimension chains. The area itself is calculated by a deterministic geometry engine – same input, same result. The confidence says how certain the detection was, not how closely the plan matches the building: sign-off and survey remain with people.