Deep Belief Neural Networks for Eye Localization Based Speeded up Robust Features and Local Binary Pattern

Eye localization and detection are essential in security applications and human recognition and verification. The multi-pose variations of the pupil are still the major challenge in eye detection algorithms. Furthermore, facial expression recognition related to eye detection is still dropped in the recent security applications. This paper used a speeded-up roust feature (SURF) algorithm to localize facial parts, especially the eye and pupil, quickly and easily. Moreover, we detect the boundary box of face components by initializing the eye position based on Hough circle transform (HCT) and local binary pattern (LBP). Afterward, we classify the individuals who successfully detected their eye images using the confusion matrix of two class labels based on deep belief neural networks (DBNN). Fine-tuning the hyper-parameter values of the DBNN is performed as well as a stochastic gradient descent optimizer to handle the overfitting problem of the proposed method. The proposed algorithm’s accuracy based on the combination of SURF, LBP with the DBNN classifier reached 95.54%, 94.07%, and 96.20% for the applied ORL, BioID, and CASIA-V5, respectively. The comparison of the proposed algorithm with the state-of-the-art is performed to indicate that the proposed algorithms are more reliable and superior.