Hull-side marking proposals often jump from a camera or LiDAR to an autonomous result. The difficult middle is less dramatic: which drawing revision defines the intended line, which measured references anchor the local frame, which samples support the reported deviation, and who is authorized to accept the result? Hull Mark Mapping Workbench v0.1.0 makes that middle layer visible.
The release is a browser-only review tool. A user describes a bounded 2D panel, pastes reference control points, defines intended line segments, and optionally adds measured sample points. The workbench draws the records in a common local frame and calculates point-to-segment RMS, 95th-percentile absolute deviation, maximum absolute deviation, and the share of samples inside a user-supplied tolerance. It exports a normalized JSON review record and a target-level CSV summary.
Why a local frame comes first
OpenDroneMap documents ground control points as a way to correct distortion and reference data to known coordinate systems. Its public workflow also makes the coordinate reference and point provenance explicit. Hull Mark Mapping Workbench borrows that evidence discipline, not the full OpenDroneMap photogrammetry engine. The current release does not process images, fit a point cloud, or reconstruct a curved hull.
The distinction matters on a ship-side surface. A coordinate that looks precise can still be wrong because the target was misidentified, the drawing revision changed, the measuring instrument was not calibrated, or the transform from image space to the working frame is invalid. The tool therefore asks for evidence references alongside points and drawing references alongside targets. It can expose an empty field; it cannot authenticate the referenced source.
What the calculations mean
For each measured point, v0.1.0 calculates the shortest Euclidean distance to its associated finite target segment. RMS is the square root of the mean squared deviation. P95 is the nearest-rank 95th percentile of absolute deviations. The within-tolerance share counts samples whose deviation is less than or equal to the tolerance entered for that target.
These statistics are descriptive screening outputs. The tolerance is not supplied by UAMKT, a class society, a shipyard, or a regulator. The bundled project is synthetic and exists only to exercise parsing, visualization, validation, and export. Its coordinates and tolerances are not field specifications or observed performance.
Mapping software is not a flight controller
ArduPilot Mission Planner documents survey-grid and camera-trigger workflows. Hull Mark Mapping Workbench deliberately stops before that control layer. It does not generate waypoints, gimbal commands, marker commands, geofences, or emergency logic. Any operational system must independently validate the aircraft, payload, sensor calibration, standoff distance, metallic and GNSS-challenged environment, rotor-wall interaction, site authorization, and EHS controls.
That boundary also protects the useful part of an open release. A procurement or engineering team can inspect how records are parsed and how deviations are calculated without receiving UAMKT proprietary control logic or mistaking a browser demonstration for a field-tested aircraft.
Regulated marks require independent authority
The IMO Load Lines amendments in resolution MSC.143(77) prescribe dimensions and positioning for the deck line and load-line mark, require contrasting and permanent marking, and withhold the certificate until an officer or surveyor certifies correct and permanent indication. For that reason, a target labelled regulated-load-line receives a warning in the workbench. The software cannot turn a coordinate comparison into statutory or class acceptance.
Fabrication reference lines, paint boundaries, draft scales, and assigned load-line marks should not be treated as one undifferentiated object. They can share an evidence workflow, but their drawings, tolerances, acceptance authority, permanence, and verification routes differ.
Source package and evaluation
The versioned Apache-2.0 source archive contains the UAMKT-authored parser, validation rules, deviation calculations, regression tests, synthetic input, README, notices, package metadata, and license. It copies no ArduPilot, PX4, OpenDroneMap, classification-society, or shipyard source code.
Evaluation should begin with the synthetic sample, then a de-identified historical panel record whose drawing revision and coordinate provenance can be independently checked. A reviewer should recalculate several point-to-segment distances, reproduce RMS and P95, inspect out-of-bounds and duplicate-ID blockers, and verify that exports preserve the source references.
The valid claim for v0.1.0 is narrow: it organizes user-supplied local-coordinate reference lines and measured samples into a transparent human-review record. It does not establish flight performance, marking performance, field safety, dimensional acceptance, statutory compliance, or commercial readiness.
