RAPID — A model of fast eye pupil registration and tracking by a modified metaheuristic differential evolution method based on the Verhulst-Pearl equation

Abstract

This paper proposes a model of fast registration and pupil tracking — «RAPID», for devices with limited computing resource (weak personal computers, smartphones, embedded systems based on ARM architecture) in order to reduce the cost of technology for individual use by people with disabilities and medical institutions. The model is based on the idea of representing the process of video oculography as a multidimensional global optimization problem and its solution by the metaheuristic method of differential evolution. The optimization problem (objective function) is formalized as a search for the region that approximates the pupil in the three-dimensional parameter space most completely — the position and approximate size of the pupil. For the considered optimization problem we propose a modification of differential evolution method based on the process of formation of genetic isolations of population of solutions in the neighborhood of all local and global extremums of the target function followed by growth of the most adapted isolation (near the global extremum) and degeneration of others according to the differential Verhulst-Pearl equation. This behavior makes the search algorithm less «greedy» and makes it possible to correctly extract the pupil from the full frame. The developed tracking model can be used in the development of software packages in the task of augmentative communication for patients with lateral amyotrophic sclerosis or diplegia syndromes, on non-specialized devices, as well as in ophthalmological complexes and infrared-pupillometers.

Keywords

• videooculography
• differential evolution
• multivariate global optimization
• region of interest
• hough transform
• verhulst-pearl model
• видеоокулография
• дифференциальная эволюция
• многомерная оптимизация
• регион интереса
• преобразование хафа
• модель ферхюльста-пирла