From known deformation to image-based measurement.
Apply a force. Track the speckles.
Compare the measurement with the mechanics.
Surface colours show analytical stress. Red marks high stress on this experiment’s scale, not material failure. Normal and shear options show absolute magnitude; signed values remain in Field comparison.
3D display of the existing 2D solution, extruded through the entered thickness. No out-of-plane deformation is calculated. Playback and magnification change only the view; DIC results stay unchanged.
One selected object, one configured experiment. No calculations run until you click Run experiment.
Click a subset in the reference image to inspect its measured match.
Each patch is matched with integer ZNCC, then refined below one pixel using a local affine transformation.
Cyan boxes show measured locations, not enlarged motion. Rejected centres appear grey. The target polygon includes the fitted local stretch, shear and rotation.
Shared scale for analytical and recovered values. Error has its own scale centred at zero.
Dots are actual valid measurement centres in undeformed coordinates. No data are invented between dots or inside the hole. Differentiating measured displacement can amplify strain and stress errors.
Run an experiment to compare fields.
Analytical values and independent DIC estimates at the same locations.
Points are actual subset centres. Lines connect samples for readability; gaps mark unavailable data or the hole. In paired runs, error metrics use common valid centres.
Same texture, same load, same settings. Only Gaussian noise changes.
Errors use the intersection of valid centres. Coverage is reported separately. “No added noise” still includes 8-bit quantization, interpolation and the selected blur/brightness. A single trial need not worsen every metric.
The line graph in Field comparison overlays ground truth and both DIC runs.
Analytical Saint-Venant torsion of a solid circular shaft. No DIC measurements.
Zero at the centre; maximum at the outer surface. Stress is uniform along the shaft under constant torque.
Two calibrated cameras. Four images. Independently recovered surface motion.
Green dots: accepted stereo matches. Grey dots: rejected candidates in the left reference image. Images show the actual load, without visual magnification. Only the common visible front face is measured.
Truth and DIC share a colour scale; signed error uses a separate scale.
Same reference centres · dashed truth and measured dots
Shear magnitude peaks near side midpoints and approaches zero at the corners. This is the square-section solution, including warping.
Surface strains come from local quadratic fits of both reconstructed 3D surfaces. Rigid rotation is removed by the strain definition. Blank estimates mean insufficient reliable support. Camera calibration is ideal; lens distortion and experimental calibration uncertainty are not simulated. A single observed face does not determine interior stress.
The images are the only motion information passed to DIC.
Analytical displacement supplies known ground truth. Each image uses the selected geometry, elastic constants and load.
Inverse warping moves a seeded speckle pattern. Blur, brightness, independent noise and 8-bit quantization form the measured images.
ZNCC locates subsets. Six-parameter affine optimization refines displacement. Invalid or weak matches are rejected.
Local polynomial fits give strain. Plane-stress elasticity gives stress. Results are compared at the same reference centres.
No experiment has run yet.