Calibration Groups and Repeated Measurements

Curve Digitalizer can keep several calibration realizations for one and the same source image, distinguishing the active measurement, the draft, accepted or rejected state, and the measurements eligible for selection in Publish.

1. One Source for Several Calibrations

The test uses a single canonical Image Dataset image.raster, a synthetic 1280 × 900 pixel PNG with a quadrilateral graph, two colored traces and a control target declared at 5 s, 30 mV. The figure is not a real experiment; it serves to reproduce the calibration clicks in a controlled way.

After selecting the Dataset and the Preview source image → Use this image dialog, the test Digitizer Configuration V09 Calibration Repeat Group was created.

Preview source image dialog for the synthetic graph used in the repeated calibration test.
Figure 1. The canonical scientific Dataset is chosen in the interface and confirmed without reusing a previous Session.

The traces in the PNG are not yet digitized canonical curves: initially the Session contains only an image and a configuration.

2. The Calibration measurements Panel

In the Calibration tab, below Projective calibration and verification, you find Calibration measurements: a table with the Measurement and State columns, single-row selection, its own scrolling and seven buttons:

New, Activate, Accept, Reject, Cancel draft, Rename and Delete.

Calibration measurements table with Measurement 1 in the Active · draft state and the seven buttons.
Figure 2. The initial configuration shows a draft realization before the four calibration points.

A new configuration normally contains Measurement 1, already selected and Active · draft. Its measurement identifier is assigned by the Viewer; the group has another, separate persistent identifier. These technical names are used to verify persistence and must not be edited manually.

3. Accept Does Not Accept an Incomplete Calibration

With the initial measurement still lacking the four corners, Accept was pressed. The GUI showed:

A complete four-corner calibration is required.

Error message stating that a complete four-corner calibration is required.
Figure 3. A measurement without four corners cannot be classified as accepted.

The Workspace state remained draft: a simple row in the table is not the same as a geometrically valid calibration. The same condition is required for a realization that will later be selectable in the Publish tab.

4. Protecting the Last Measurement: Cancel draft and Delete

Two independent safeguards are shown on a configuration with a single measurement.

Cancel draft responded:

At least one measurement must remain.

Message stating that at least one measurement must remain.
Figure 4. The Viewer prevents the configuration from being left without a calibration measurement.

Delete showed:

At least one calibration measurement must remain.

Message stating that at least one calibration measurement must remain.
Figure 5. The Delete safeguard prevents deleting the only realization.

In neither case did the Viewer leave an empty group. The first measurement remained present, with the same ID and the draft state.

5. Setting the Shared Scientific Axes

A linear scale was set with X Elapsed time, unit s, limits 0–10 and Y Signal, unit mV, limits −20–80. The data were applied with Apply axis settings.

Axis settings with Elapsed time in s from 0 to 10 and Signal in mV from −20 to 80.
Figure 6. The scientific axis fields are shared as the basis for repeated calibrations.

The geometry of each realization is then recorded through Set BL, Set BR, Set TL and Set TR. When a new measurement is created, the limits and units already defined remain available as the basis for the new survey.

6. First Calibration: Four Real Corners

For the first realization the corners were acquired on the raster, using the four buttons and actual clicks on the canvas. The GUI showed the cyan markers BL · M1, BR · M1, TL · M1, TR · M1 and the status Calibration ready.

Four cyan markers BL, BR, TL and TR · M1 on the graph with Calibration ready.
Figure 7. The four corners of the first realization were acquired with clicks on the canvas.

The four points represent the correspondence between source pixels and the normalized corners (0,0), (1,0), (0,1), (1,1) of the scientific surface. The transformation is projective, even though the numerical axes are linear.

7. Accepting Measurement 1

After the complete calibration, Accept changed the row from Active · draft to Active · accepted.

Calibration measurements table with Measurement 1 in the Active · accepted state.
Figure 8. Acceptance changes the state of the measurement while keeping the same geometry and source.

The framework keeps the same measurement_id, calibration_group_id, coordinates of the four corners and axes. Acceptance expresses an operational decision on the validity of the realization; it does not automatically introduce numerical uncertainty, nor does it publish Datasets.

8. New: A Second Measurement Without Deleting the First

Pressing New, the Viewer created Measurement 2, automatically activating it as draft. The current four markers were reset, while Measurement 1 remained accepted and kept in the same group.

Table with Measurement 1 accepted and Measurement 2 active as draft, with no markers on the graph.
Figure 9. New keeps M1 accepted but opens a second survey with zero corners.

The Workspace therefore has two distinct IDs: M1 still contained its four points, M2 contains none. The scientific axis settings are inherited.

New does not delete raster points already digitized.

9. Selecting a Row Does Not Mean Activate

With M2 still active but incomplete, only the accepted row Measurement 1 was selected. The table highlighted M1; however, the text Corner markers — ACTIVE still indicated M2.

Table with Measurement 1 highlighted while Corner markers — ACTIVE still indicates Measurement 2.
Figure 10. The highlighted selection does not automatically change the active scientific mapping.

Verify calibration… was then pressed. The viewer refused the verification because the calibration was incomplete:

Complete the four-corner calibration and valid scientific axes before verifying a known point.

Warning asking to complete the four-corner calibration before verifying a known point.
Figure 11. Verify calibration uses the Active measurement, not the merely highlighted row.

This matters: Verify calibration uses the active realization, not simply the highlighted row. The selection and activation functions are deliberately different.

10. Activate: Restoring the First Transformation

With M1 selected, pressing Activate brought back the four markers of the first calibration; the accepted measurement became active again and the second one remained saved as a draft without points.

Measurement 1 activated again with its four calibration markers visible.
Figure 12. Activate restores the original geometry without recreating the raster.

The interface did not recreate the image, and the CORE reloaded exactly the corners already persisted. To continue the second measurement, simply select M2 and press Activate.

11. A Second Realization with Slightly Shifted Corners

On the same image, M2 was completed with four new clicks slightly different from those of the first realization. Instead of assuming the same transformation, the Viewer kept a second, genuinely distinct homography.

Four calibration markers of Measurement 2 slightly shifted from those of Measurement 1.
Figure 13. The four clicks of M2 create a different homography on the same Image Dataset.

The group thus had two complete realizations, separated by ID and point coordinates but with the same X/Y ranges.

12. Two accepted Calibrations of the Same Raster

With Accept, the second measurement was approved as well. The table showed Measurement 1: accepted and Measurement 2: Active · accepted.

Table with Measurement 1 accepted and Measurement 2 Active · accepted.
Figure 14. Two distinct realizations are kept in the same calibration group.

The difference between the two transformations is demonstrated numerically on an independent point later in the chapter. Accepting two calibrations does not mean claiming that the two surveys are two independent physical experiments.

13. Rename Keeps the Scientific Identity

The Rename command on the second row opened the native Rename calibration dialog, with the Measurement name field.

Rename calibration dialog with the Measurement name field.
Figure 15. The native prompt changes the label without assigning a new scientific ID.

The new name assigned is Repeat B – shifted corners. The table and the text of the active realization were updated.

Table showing the second measurement renamed Repeat B - shifted corners.
Figure 16. The updated description still refers to the second realization.

14. Third Realization: Complete Calibration but rejected

A third measurement was created and four new corners were selected, again on the same image, with the same pair of linear axes.

Four calibration markers of the third measurement on the same graph.
Figure 17. A third complete realization has coordinates slightly different from M1 and M2.

The Reject command changed the state to Active · rejected.

Table with the third measurement in the Active · rejected state.
Figure 18. Reject changes the operational state but keeps the ID and corners of the realization.

The rejected measurement remained in the group, with its ID and four corners. Reject is not the same as Delete. Its transformation can still be computed mathematically, but by default it does not belong to the set of measurements eligible for publishing.

15. Cancel draft Works Only on Drafts

Pressing Cancel draft while the rejected measurement was selected, the Viewer showed:

Only draft measurements can be cancelled. Use Delete for an accepted measurement.

Message stating that only draft measurements can be cancelled.
Figure 19. Cancelling drafts only does not remove a rejected measurement.

The same behavior occurs on a measurement that is already accepted. The interface prevents a button designed to discard drafts from accidentally destroying realizations in other states.

16. Only Accepted Measurements Are Eligible in Publish

Switching to the Publish tab, the Select accepted calibration measurements section contained exactly two rows: Measurement 1 and Repeat B – shifted corners. The third, rejected measurement does not appear, even when it is the active measurement in the Calibration tab.

Publish tab listing only the two accepted calibration measurements.
Figure 20. The choice for publishing is filtered on the accepted state, regardless of Active.

This confirms a second essential distinction: the active measurement determines the current mapping in the Viewer, while the selection for publishing admits only accepted realizations.

The GUI offers separate and combined in the Publication layout field; the numerical structures, the calibration_key, the columns and the scientific Preview will be verified in the dedicated chapters.

17. The Accept/Reject Decision Does Not Force Activation

After activating M1, the third, already calibrated measurement was selected. Accept moved M3 from rejected to accepted while leaving M1 active.

Table with Measurement 3 accepted while Measurement 1 stays active.
Figure 21. Accept changes the state of the selected row without activating its transformation.

A subsequent Reject restores the rejected state, again without changing the active measurement or the stored calibration points.

Table with Measurement 3 back in the rejected state.
Figure 22. Reject lets you revise the classification while keeping the four points.

It is therefore possible to review the classification of a realization without silently swapping the scientific transformation in use.

18. Cancel draft with No/Yes Confirmation

To verify discarding drafts only, a fourth temporary measurement, Measurement 4, was created, active and with no new corners.

Table with a temporary Measurement 4 active as draft.
Figure 23. The temporary measurement is distinct from the three realizations already preserved.

Pressing Cancel draft, the Viewer asked for explicit confirmation:

Discard draft ‘Measurement 4’?

Confirmation dialog asking to discard draft Measurement 4.
Figure 24. Cancel draft requires an explicit decision before discarding the draft.

Choosing No kept all four measurements. Repeating the command and choosing Yes, only the M4 draft was removed; the program reactivated M1, accepted and available.

Table after discarding Measurement 4, with two accepted and one rejected measurement.
Figure 25. The M4 draft is removed and the group keeps two accepted and one rejected.

The two accepted measurements and the rejected measurement remained unchanged. No canonical source was deleted.

19. Delete: Confirmation and Persistence of the Group

Another fourth temporary measurement was created to show Delete separately. The command showed:

Delete Measurement 4 permanently?

Confirmation dialog asking to delete Measurement 4 permanently.
Figure 26. Delete requires a confirmation separate from the Cancel draft path.

Choosing No left the measurement in place. With a second request and Yes, the Viewer removed only M4, preserving the group and the three previous realizations.

Table after deleting Measurement 4, with the three previous realizations.
Figure 27. The targeted removal does not change the IDs of the existing realizations.

20. Same Pixel, Two Scientific Coordinates

To compare M1 and M2, the same point of the raster was clicked at the identical canvas coordinates (367,379). The resulting source pixel was about (675.9137; 445.3957).

With Measurement 1 active and the known reference Expected X=5 s, Expected Y=30 mV, Verify calibration… returned:

Calibration verification with Measurement 1 active and Expected values 5 s and 30 mV.
Figure 28. The panel uses the first homography and reports the residuals with respect to 5 s and 30 mV.

Repeating the same click with Repeat B – shifted corners active:

Calibration verification with Repeat B - shifted corners active on the same clicked point.
Figure 29. The shifted M2 transformation produces different scientific coordinates on the same pixels.
Accepted measurementCalculated X (s)Calculated Y (mV)
Measurement 14.99186930.10595
Repeat B – shifted corners5.03065030.26146
Difference M2−M1+0.038781+0.155509

The second realization changes the pixel → science mapping, not the PNG pixels.

The differences are not automatically a standard deviation or a metrological uncertainty: they are two evaluations of the same raster with different calibration pointing.

21. Selecting the First Row with the Second Measurement Active

At the end of the numerical comparison, Measurement 1 was selected in the table without pressing Activate. The Viewer kept showing the cyan M2 markers and the text ACTIVE: Repeat B – shifted corners.

Table with Measurement 1 selected while the markers and ACTIVE text refer to Repeat B - shifted corners.
Figure 30. The table and the markers still distinguish selection and activation after acceptance.

The numerical verification and the display of the markers depend on the active realization, while some table actions, such as Rename, Accept, Reject and Delete, operate on the selected realization. Always check the ACTIVE text before interpreting data or verifications.

22. Selection for Publishing and Two Possible Layouts

In the Publish tab, Select all accepted was pressed. The two eligible rows were selected, while the rejected one stayed excluded.

Publish tab with both accepted measurements selected and the Publication layout choice.
Figure 31. Select all accepted does not include the third, rejected measurement and does not produce Datasets.

The GUI shows the choice Publication layout: separate / combined. The interface describes separate as one Dataset per calibration, and combined as a single Dataset with a column identifying the calibration. In V09 this is an observation of the presentation plan: Publish selected calibrations was not invoked and there are no new canonical tables to compare.

23. The Complete Table After Scrolling

The Calibration tab is scrollable. Scrolling it down, the Calibration measurements table shows all three rows with their final states and the whole bar of seven buttons.

Scrolled Calibration measurements table with three rows and the seven buttons.
Figure 32. The scrollable table shows two accepted, one rejected and the whole New/Activate/Accept/Reject/Cancel draft/Rename/Delete bar.

The calibration frame and the measurement table have different functions: the first manages the corners of the active realization, while the second determines which measurement is selected, activated, classified or removed. The frame also has its own independent internal scrolling.

24. Save, Close and Reopen in the Same Session

The final configuration was saved with Configuration → Save configuration, closed through the normal window protocol and reopened from the Viewers → Digitalizer catalog.

Reopened Viewer with the calibration group and its three realizations restored.
Figure 33. After Save and Close/Reopen in the same Session, IDs, transformations and states match the saved Workspace.

The framework contains a single canonical image, a single calibration group and the three final realizations:

LabelStateStable IDCorners
Measurement 1acceptedDistinct4
Repeat B – shifted cornersActive · acceptedDistinct, preserved after Rename4
Measurement 3rejectedDistinct4

25. What Repeated Measurements Mean Scientifically

The three realizations are different choices of the four calibration points of the same digital graph. They are not independent experiments performed with a different instrument or sample.

The group keeps the IDs, transformations, axes and states. According to the installed model, the measurement ID is distinct from any fingerprint of the transformation; two geometrically identical mappings can belong to measurements with different IDs.

No variance, standard deviation, covariance or physical acquisition errors are derived automatically from these calibrations. It is possible to study the sensitivity of the calculated values to corner pointing, but a metrological assessment requires additional, declared assumptions and tools.

26. Short Operating Procedure

To study several calibrations of the same raster: define the axes, acquire four corners for M1 and check the geometry; press Accept. Use New to create M2, select four corners and accept it if suitable; repeat for further realizations.

Always check the difference between the selected row and the ACTIVE measurement. Use Activate when you want to change the current mapping and Verify calibration… to compare a known point. Reject keeps an unsuitable measurement; Cancel draft removes only a draft, with confirmation; Delete requires confirmation and cannot leave the group empty.

The Publish tab admits only accepted realizations, but the creation and structure of the Datasets will be verified in a later guide.