Appendix

Glossary

Terms as this manual uses them. Where SurView's own screens use a different word for the same idea, that word is given alongside it.

Reference image

The photograph taken before deformation. Every displacement in a run is measured relative to it.

Target image

A photograph taken after some deformation has occurred, compared against the reference (or, after reference updating, against a previous target) to measure movement.

Subset

The small patch of pixels around one measurement point that is matched between the reference and the target. Its size is the subset radius.

Grid step

The spacing, in pixels, between neighbouring measurement points.

Correlation coefficient (ZNCC)

A number stating how closely a candidate match in the target resembles the original subset in the reference. A measure of matching confidence, not of physical correctness.

Ground

Part of the scene that did not move between two sessions, such as the bench or a fixture, marked on the reference so that a target from the second session can be aligned on it. Not measured.

Known length

Two points whose distance did not change between two sessions, such as the ends of a scale bar. Where no ground is marked, the camera's scale is taken from it.

Perspective (camera tilted)

A camera move between two sessions that includes a tilt, fitted as a homography on stationary ground, so the camera's scale varies across the picture. Chapter 14.

Displacement field

The measured movement, in pixels, at every point of the grid.

Strain

Local stretching or shearing, fitted from the displacements of neighbouring points rather than measured directly at any one point. Reported in three common forms: Cauchy (small-strain), Green-Lagrange, and Euler-Almansi, which agree closely at small deformations and diverge at large ones.

Region of interest (ROI)

The boundary restricting where measurement points are placed, drawn by hand or proposed by automatic segmentation of the speckle quality. May exclude one or more holes: places inside the outer boundary that are not to be measured.

Sequence

A reference image plus an ordered series of target images, making up one loading test.

Reference updating (re-anchoring)

Switching a sequence's comparison from the original reference to a more recent frame, once too little of the field still correlates against the original. Off by default.

Recovery pass ("Points that failed" in the Analysis panel)

A second attempt at points the first correlation could not measure well, seeded with a displacement fitted from nearby reliable points and then fully re-solved. On by default; never accepts an answer worse than the one it would replace.

Noise floor (sigma)

The finest displacement a given subset's own speckle quality could resolve against the camera's sensor noise. A lower bound, computed from the reference image alone.

Match conditioning (beta)

How sharply the matching cost rises around the solution actually found for one point. Meaningful only within a single run.

Clipped share of subset

How much of a point's reference subset sits at the darkest or brightest value the image holds, where the camera recorded no pattern. More than half is reported as a warning.

Virtual extensometer

Two points placed on the measured field whose changing distance apart is plotted against frame, standing in for a physical clip gauge.

Where to find it on screen

What is not in this manual yet

Written up only where the screen it describes actually exists, this manual is necessarily behind SurView's own roadmap. As of this writing, the following are real, tracked work that simply has not landed:

• Stereo and out-of-plane (2.5D) measurement, and the camera calibration workflow it depends on.
• Multi-view measurement from a single moved camera, for slow or static tests such as creep.
• Sparse point-marker tracking, alongside the dense speckle field.

See ROADMAP.md for the current state of each. This manual will gain a chapter for each one as it ships.