Finding and tracking Bragg spots in GISAXS maps of block Co-Polymer thin films: using cascade based feature extraction and circular Hough transformation

Real time restructuring behaviour of ~100nm spin-coated nanostructured ABC miktoarm star terpolymer films during solvent vapor annealing was monitored using GISAXS. Throughout the investigation, a large quantity of 2D GISAXS maps was produced. Some of these maps exhibit Bragg rods or Bragg spots and other features, which changes position, shape and intensity in accord with structural changes in the film. For the purpose of identifying and tracking Bragg spots with changing position, two different feature extraction algorithms have been tested: 1) a cascade based algorithm based on the Viola-Jones algorithm [1] and 2) an approach based on the Circular Hough Transform [2]. The cascade based algorithm is able to identify the Bragg spots in the dataset tested, consisting of ~3000 maps, with very few false positives or false negatives (< 10) at a rate of ~18 frames per second. The Circular Hough Transform is generally a faster algorithm (~70 frames per second) but fails for part of the dataset, during the drying stage of vapor annealing, due to changes in the shape of the Bragg spots tracked.