Robust Face Detection and Identification using HOG-based Features and Machine Learning

This paper focuses on integrating face recognition into educational websites. The method involves webcam data collection from three subjects, followed by preprocessing steps like resizing, grayscale conversion, histogram equalization, and median blur. The Haar Cascade Classifier aids real-time face detection, while the Histogram of Oriented Gradients (HOG) facilitates feature extraction, yielding around 15,000 features per image. Principal Component Analysis (PCA) is applied for dimensionality reduction. The study evaluates classifiers: Logistic Regression, K-Nearest Neighbors (KNN), Random Forest (RF), and Support Vector Machine (SVM), with Logistic Regression achieving 99.54% accuracy. This comprehensive approach enhances website security and user experience, showcasing the synergy of preprocessing, advanced feature extraction, and effective classification for robust face recognition. Future prospects involve deep learning integration, real-time processing, and adaptability enhancement.