Quality Assurance and Reproducible Workflows in Agisoft Metashape

Define why the project exists and who is responsible for each decision.

• Record the project objective, location, date range, customer or internal owner and intended use.
• List required outputs, coordinate systems, formats, resolution and accuracy expectations.
• Identify project manager, processing operator, technical reviewer and release authority.
• Define acceptance criteria before final results are generated.
• Record confidentiality, data-protection, retention and publication constraints.
• Identify applicable survey, engineering, scientific, contractual or regulatory requirements.
• Define how deviations, changes and unresolved issues will be approved.
• Assign a unique project identifier and naming convention.

Controlled record: approved project brief, responsibility matrix and acceptance specification.

Preserve what was received before files are renamed, converted or excluded.

• Inventory image folders, file counts, formats, dimensions, sensors and acquisition dates.
• Record original filenames, folder structure and source location.
• Preserve metadata relevant to camera, GNSS, timestamps and orientation.
• List GCPs, checkpoints, trajectories, calibration files, masks, laser scans and supporting documents.
• Use checksums or another controlled method where file-integrity verification is required.
• Document corrupted, duplicate, missing, converted or deliberately excluded files.
• Retain an immutable or access-controlled copy of original inputs where policy permits.
• Record data ownership, permissions and restrictions on use or publication.

Controlled record: source-data manifest, integrity record and exception list.

A result may depend on the environment in which it was produced.

• Record Agisoft Metashape edition, exact version and build where available.
• Record operating system, relevant updates, CPU, GPU, VRAM and system RAM.
• Record GPU driver and relevant Metashape Preferences settings.
• Identify local, network, server, cloud or hybrid processing architecture.
• Record scripting, plugins, Python modules or automation used.
• Identify storage location, project format and backup state.
• Preserve processing logs required for troubleshooting and audit.
• Record environment changes made during the project.

Controlled record: software and infrastructure manifest linked to the released project version.

Geospatial and camera-model decisions require explicit documentation.

• Record source, project and output coordinate reference systems.
• Document horizontal datum, vertical datum, geoid, units and transformations.
• Record camera groups, sensor models, calibration source and fixed or adjusted parameters.
• Identify camera-position source and assigned accuracy.
• Separate GCP, checkpoint and scale-bar roles.
• Record marker coordinates, accuracies, projection review and excluded observations.
• Document RTK or PPK processing and any lever-arm or antenna-offset assumptions.
• Preserve the reference configuration used for the approved optimization.

Controlled record: CRS, calibration and reference-data configuration with roles and uncertainty stated.

Record the parameters that materially define each processing stage.

• Define the planned order of alignment, optimization, reconstruction and export stages.
• Record matching accuracy, preselection, limits and relevant alignment options.
• Record depth-map quality, filtering and source choices for derived products.
• Record point-cloud, mesh, DEM, orthomosaic, tiled-model and texture parameters as applicable.
• Identify region, boundary, resolution, projection and interpolation decisions.
• Save batch-processing configurations or scripts where they form part of the controlled workflow.
• Record reused intermediate products and the project state from which they originated.
• Distinguish planned parameters from the settings actually executed.

Controlled record: executed workflow sheet, batch configuration or equivalent parameter record.

Do not allow an unresolved early-stage problem to propagate into expensive outputs.

• Define input, alignment, reference, reconstruction and release quality gates.
• Specify the evidence required to approve each gate.
• Review unaligned cameras, coverage, tie-point distribution and weak geometry.
• Review calibration behaviour, reference residuals and independent checkpoints.
• Inspect holes, noise, classification, seamlines, artefacts and output completeness.
• Record approve, revise or reject decisions with responsible person and date.
• Prevent later stages from overwriting the last approved state without a new version.
• Document exceptions and compensating controls.

Controlled record: signed or attributable quality-gate checklist with evidence references.

Manual expertise is valuable, but its effect should remain visible.

• Record image exclusions, masks, marker corrections and camera enable/disable decisions.
• Document tie-point cleaning, point-cloud edits and classification changes.
• Record mesh editing, hole filling, orthomosaic seamline edits and manual surface corrections.
• Identify the operator, date, reason and affected project state.
• Preserve a pre-edit state for changes that are difficult to reverse.
• Separate corrective edits from cosmetic presentation changes.
• Review whether manual intervention changes accuracy, completeness or uncertainty.
• Require additional approval for deviations from the planned workflow.

Controlled record: change log connecting each material intervention with its purpose and project version.

Acceptance must be connected to evidence rather than visual confidence alone.

• Generate and retain the relevant Metashape processing report for the approved state.
• Record control and independent checkpoint results separately.
• Review reprojection, camera, marker and scale-bar information as applicable.
• Include coverage, resolution, density, model statistics or other project-specific indicators.
• Compare results with predefined acceptance criteria.
• Document outliers, excluded evidence and the technical rationale.
• Identify tests conducted outside Metashape, including GIS, CAD, survey or laboratory checks.
• State limitations, untested areas and conclusions that cannot be generalized.

Controlled record: validation package containing reports, statistics, external checks, decisions and limitations.

The delivered files should be identifiable without opening the working project.

• List every delivered file with version, format, size and purpose.
• Record export coordinate system, vertical reference, units, resolution and compression.
• Identify the source project state and date of export.
• Use consistent filenames and prevent draft outputs from being mistaken for released products.
• Include processing report, validation statement, metadata and usage limitations where required.
• Verify that files open correctly in the intended recipient environment.
• Record secure transfer method, recipient and delivery confirmation.
• Protect the released package from undocumented modification.

Controlled record: release manifest and approval connecting deliverables to the validated project state.

An archive is useful only if it can be understood and restored.

• Define which source data, project files, reports, scripts, logs and deliverables must be retained.
• Confirm whether the selected Metashape project format and associated directories are archived correctly.
• Record software version and any installation resources needed for future access.
• Preserve custom coordinate systems, geoids, calibration files and external dependencies.
• Store the archive in accordance with backup, security and retention policies.
• Test opening or restoring the package from a separate location.
• Verify that links to source imagery and supporting files remain valid.
• Record archive date, custodian, retention period and authorized deletion process.
• Where required, reproduce a representative stage or export to confirm the record is sufficient.

Controlled record: verified archive package, restoration result and retention responsibility.