Implementation of the human eye pupil detection system on FPGA

Abstract

The pupil detection technique is advantageous in several real-time systems, including ophthalmic testing equipment and wheelchair assistance. Due to the difficulties associated with the variable size of the pupil, occlusion of the eyelids, and eyelashes, the pupil detection system is extremely difficult to implement over a wide variety of datasets. Over the previous decade, numerous methods have been developed to detect the pupil. These algorithms affect non-contact tonometers, automatic refractor keratometers, and optical coherence tomography. The standard technique of pupil detection employs an algorithm based on conventional neural networks that additionally include an integrodifferential operator and a circular hough transform. It is a disadvantage because it is inefficient for applications that require extensive logic resources to implement. To address this limitation, this research effort proposes a new pupil identification approach based on an average black pixel density that accurately recognizes the human pupil region with minimal logic resources in an FPGA. Along with average black pixel density, double threshold, and logical OR, this suggested work incorporates morphological closing modules to increase the sensitivity, specificity, and accuracy of the proposed work. newline