Year
2018
Abstract
The Zero-Knowledge Protocol (ZKP) for warhead verification, developed by Glaser, Barak and Goldston, and experimentally validated by Philippe, Goldston, Glaser and d’Errico, depends on an analog means to store measurements of neutron fluence. Experiments to date have focused on super-heated droplet, or “bubble” detectors, developed by d’Errico. We have found that conventional bubble-counting techniques, based on a limited number of photographic images, saturate vs. fluence due to the occultation of bubbles by one another. However the efficiency and/or effectiveness of the ZKP technique depends on the quality of the statistics that are obtained in these measurements. The effect of occultation can be approximated by an analytic expression that is easily inverted to estimate the true number of bubbles. Once calibrated, this approach works well, but the uncertainty will increase at high fluence. We are now investigating the use of movies of rotating bubble detectors. Such movies clearly show bubbles emerging from behind one another, and in principle should allow significantly more accurate bubble counts at high bubble density. The Hough transform, a machine vision technique based on a “voting” algorithm, has been used successfully to interpret bubble-chamber images and to study the interstellar medium. Here we investigate using this transform - twice - to interpret bubble detector movies. First, we use the well-known Elliptical Hough transform to find the bubbles in each image. In effect each pixel in a detected bubble edge votes for all ellipses that could contain it. Those ellipses that obtain the most votes are deemed to correspond to bubbles. Next, in a new application of the Hough transform concept, the bubbles in each movie frame vote for all possible rotational curves that could contain them. This allows bubbles that emerge for only a fraction of the movie frames to be detected. The accuracy of this technique will be validated against experimental data.