ENGINEERING PROFAS / Evidence models
Pixels, surfaces and signals
The same examination record can connect several kinds of evidence. The evidence itself still needs its own coordinates, units and analytical limits.
In this article
Share the record envelope
The original centre of PROFAS was image comparison: place observations on material, explain them and preserve the basis of an evaluation. Extending that idea to measured surfaces and DNA traces creates an architectural question. Which parts of the application describe examination, and which parts describe the evidence itself?
Authorship, acceptance time, exact revisions and citations belong to the examination record. A contribution develops within a stage; a submitted basis remains fixed; independent review has its own baseline and disclosure sequence. Those responsibilities remain useful whether an observation refers to a ridge, a surface region or a signal interval.
The module definition supplies specialist tools, assessment vocabulary, permitted opinions and measurement methods. Definitions are versioned in the application, and examinations retain their selected version. This is a maintained set of explicit contracts rather than an unrestricted plug-in loader where arbitrary code can redefine what an existing record means.
The common envelope therefore creates continuity without forcing all content into the same payload. A source-linked image annotation, a surface measurement and a DNA observation retain distinct representations. Their shared lifecycle makes them inspectable together; their differences prevent the interface from suggesting analytical equivalence.
Three native frames, one contribution lifecycle
- Image pixels
Original capture coordinates; physical distance requires an explicit calibration basis.
- Measured surface
Integer native grid samples, source-declared physical axes and retained missing values.
- DNA signal
Channel and native scan positions, with separately identified imported peak analysis.
Keep image geometry in original pixels
A fingerprint may be displayed at several zoom levels or alongside a reference of different dimensions. Its annotation should continue to identify the same source location. PROFAS therefore separates the coordinates of accepted geometry from the transformation used to display that geometry in a pane.
The browser maps a pointer through the current source transformation. Agent tools can name either the original capture frame or a registered returned-image frame. The accepted result refers to the source rather than the viewport, so changing the screen arrangement does not redefine the selected location.
Units need equal care. Two points provide a distance in pixels, but an image does not acquire physical scale because the interface draws a ruler. A calibrated measurement depends on the exact calibration revision and source to which it applies. Uncalibrated distance retains its pixel unit instead of implying millimetres.
The surrounding interpretation also remains specialist. A fingerprint quality class is not a universal confidence score that can be copied onto a tyre region or DNA peak. The module validators preserve these distinctions, allowing the common workspace to support different disciplines without flattening their vocabulary.
Treat a surface as measured samples
The measured-surface implementation began with a pair of NIST X3P cartridge-case scans. The supported import is a bounded height-grid representation: incremental X and Y axes with absolute numerical Z values. The importer converts its supported SI metre representation into micrometres and retains source and canonical identities.
A shaded mesh is useful for inspection, but it is a presentation of those samples. Lighting, display reduction and vertical exaggeration can make marks easier to see. They cannot become inputs to the accepted numerical measurement. The measurement functions use integer native-grid sample locations and the source-declared axes.
Missing samples remain missing. Measurement endpoints must refer to measured values, while a sampled profile retains gaps along its path. Filling holes for a smoother picture would alter the analytical basis, so the presentation and measurement code cannot quietly substitute interpolated values for absent data.
The units have limits too. Source pitch does not establish measurement uncertainty, and two scans do not automatically share a meaningful Z origin. A signed height difference on one grid is different from subtracting arbitrary elevations across separate acquisitions. Preserving these constraints makes the result more useful to inspect than an unexplained number.
Treat a trace as a signal
DNA STR interpretation introduces another frame. The selected PROVEDIt sources provide raw numeric detector channels in native scan indices. Their associated analysis exports provide locus rows, allele designations, fragment sizes and peak heights separately. Both are useful, but they do not supply the same kind of information.
In particular, the imported peak table does not establish a continuous sizing curve or a position on the native trace for every peak. PROFAS presents the scan-domain signal and separately identified imported peaks. It does not place a supplied allele designation on the trace by guessing where it belongs.
An observation on the native signal names its channel and sample or interval. An interpretation of a supplied peak names the exact imported peak identity. Each carries its source frame and explanation. This lets another reader distinguish the recorded interpretation from the values imported with the source.
The same boundary extends to statistical evaluation. PROFAS records whether statistical work was not performed or supplied externally, retaining the external report and its stated basis. It does not turn a comparative display into a probabilistic-genotyping engine or calculate a mixture likelihood ratio from these observations.
Reject operations that do not fit the source
Extensibility becomes more credible when incompatible operations fail clearly. DNA records require their exact signal frame; raster and surface record types are rejected for that module. Surface measurements require the surface-capable release, permitted method and native frame. An image transform cannot stand in for either analytical source.
Adding a modality consequently involves real work: a supported parser, canonical representation, coordinate contracts, views, validation, recording and extraction. The shared workflow removes the need to invent authorship or review history again. It does not remove the need to understand the new evidence or preserve its limitations.
That trade-off is intentional. A universal canvas object would be simpler to draw and harder to interpret. Explicit payloads and frames require more deliberate implementation, but they give a later reader enough information to ask which source, which units and which operation produced a result.
The current modules remain Development implementations. Supported files and technical rehearsals establish particular software behaviour, not universal format compatibility, specialist method acceptance or scientific performance. The architectural aim is a common record that can accommodate those separate decisions while keeping the material and its interpretation connected.