next up previous contents
Next: Recommended settings Up: Model Based calibration Previous: Model Based calibration   Contents

Model Parameters

Feature No; specifies a single feature from the wire-frame model for display, a value of zero selects all features, and negative numbers identify occluding boundaries.

Starting from a close initial transformation, the "Model min" routine will adjust for the best alignment in both images. Debug plots for sample distributions of Likelihoods and hypothesis tests are made available for viewing in the Histogram Tool.

"gradient/variance" options specify the form of edge model used. The former being a step edge and the latter local image variance computed as a difference of Gaussians.

The "edge/orient" choice menu specifies the hypothesis and likelihood terms.

Optimisation is determined by minimising a cost function in three parts which are separately weighted by the; scale edge, scale angle and scale shift parameters. The theoretical value for each is 1. Other factors are:

Shift factor (1.0); base lateral shifting, as specified in ffgnn files, is multiplied by this factor when applied up to a maximum shift of 15 pixels.

pixSep (2.0); mean sample rate in pixels along lines and ellipses.

pixDiv (1); samples per pixel in lateral shift direction.

OrrErr (0.02); Quantitation of the image into pixels restricts orientation accuracy on measured points. This parameter specifies this limiting accuracy.

Hypothesis verification (not aliment) is controlled by;

hyp_limit (0.01); The test for point detection.

% thresh; specifies the proportion of a curve which must be present in order for it to be displayed.


next up previous contents
Next: Recommended settings Up: Model Based calibration Previous: Model Based calibration   Contents
root 2017-09-26