Before data is processed further, the AI checks its suitability for the intended purpose. Typical quality issues often arise in the field: blurred or skewed image strips, underexposure or overexposure, insufficient GCP distribution, RTK failures, as well as occluded areas or abrupt changes in elevation.

This identifies faulty datasets early, before compute time, manual rework, or repeat field operations are required.

A key added value lies in the automated derivation of CAD- and GIS-ready features from point clouds, DSM/DTM, and orthobased elevation models. The AI detects edge contours, evaluates normal vectors, and segments surfaces by slope, curvature, and surface structure. This turns unstructured data into reliable CAD objects.

For construction, landfills, quarries, recycling, and earthworks, we automate the calculation of quantities and changes. This makes volumes faster, repeatable, and cleanly documentable - with versioning, parameterization, and reproducible evaluation.

With repeated measurements, the AI detects changes between two or more points in time. It combines geometric differences with semantic classification and thus delivers not only a difference map, but also a domain-specific statement: what has changed, where it happened, and whether it is relevant.

The DoD approach (DEM of Difference) is the core of the analysis: epochs are co-registered, then differentiated epoch by epoch. A threshold filter separates significant changes from noise and artifacts; uncertainty estimation takes into account point density, GSD, local roughness, interpolation errors, and registration residuals.

The semantic component focuses on domain-relevant classes such as “stockpile” or “ground”. Vehicles, machinery, vegetation, or temporary disturbances can be excluded or treated separately depending on the use case.

With more than two epochs, the before-and-after comparison becomes time series analysis. This makes it possible to monitor trend directions, rates of change, stability phases, recurring patterns, and step changes over longer periods.

We compare surveying data with design data and output deviations in a structured way. Tolerance rules by trade, area, or component create clear decision-making foundations for acceptance and construction processes.

Tolerance rule logic: Each measurement is checked against a rule-based threshold matrix. Example: ±20 mm for concrete surfaces, fair-faced concrete, or precise installed components; ±50 mm for earthworks, subgrade, or non-finish structural surfaces. The rules can differ by component type, execution phase, zone, measurement resolution, or material class. If the permissible deviation is exceeded, the system generates a notification with severity and a reference to the affected rule.

Our AI solution does not replace existing tools, but instead complements surveying, CAD, GIS, and BIM processes as an AI layer. It automates analyses, checks, classifications, and handoffs — vendor-independent and without media breaks.

The integration architecture supports REST APIs for requests, webhooks for events, batch processing for large data volumes, as well as cloud-native deployments in hybrid or fully cloud-based environments.

Platform and interface integration: Pix4D, Metashape, Civil 3D, BricsCAD, QGIS, ArcGIS, Procore, ACC, and Jira are connected via API, webhooks, file exchange, object IDs, and structured exports.

We do not replace surveying engineers, civil engineers, or specialist reviewers. The AI handles preprocessing, pattern recognition, prioritization, and preliminary review — including data normalization, object/feature recognition, plausibility checks, deviation detection, volume and area calculations, and reproducible audit logs. Technical assessment, correction, approval, and liability remain with the expert.

Customer data remains within the agreed operating environment and is not used for external training purposes. Models, training data, parameters, configurations, prompts, outputs, and audit logs are versioned, documented in an audit-proof manner, and controlled through defined approval and rollback mechanisms. We deliver the complete technical operating foundation — from public cloud to isolated on-premise environments.

We turn surveying data into not a black box, but a traceable, integrated, and acceptance-ready technical process chain. All results are documented, versioned, and directly integrable into existing workflows.