Desarrollo de un prototipo móvil para identificar contorno de ojos en imágenes de rostros humanos utilizando programación en paralelo para un sistema de autenticación biométrica

Abstract

Biometric authentication in simple words is the identity verification of people through cha racteristics that are unique in the human body, developing this type of authentication system re quires a large computational load, being a big problem at the time of development. The present work focuses on the detection of the eye contour of the human face in an efficient and optimal way through light programming languages such as C / C ++ applying parallel programming techniques and computer vision. The methodology used follows the steps of reading face ima ges which will go through the DLIB shapepredictor pre-trained model, which identifies human faces with the help of 68 facelandmarks, identifying reference points on the face, then cutting out the eyes. and runs them through a medium hybrid filter to remove salt-and-pepper noise then specular highlight detection to remove eye glare. A data set of 500 images of faces and eyes from the CASIAWebFace repository and two own datasets with 117 and 593 images were used. Parallel programming techniques are applied that help optimize available computational resources, since by taking advantage of all the capacity of the device, waiting times between processes can be shortened by 20% and CPU consumption by 35% due to manipulation. of repetitive tasks that can be executed simultaneously without affecting other processes with the help of OPM policy synchronization.