Manual attendance methods such as paper register are error-prone and proxy attendance in educational institutions. This research proposes an automated, real-time attendance system that employs face recognition to alleviate these problems. The system uses Haar Cascade classifiers for face detection and the Local Binary Pattern Histogram (LBPH) algorithm for face recognition, processing real-time classroom video streams. The system was trained and tested using a database of 3600 facial images taken under different lighting and pose conditions. The model achieved 95% accuracy in normal lighting, 91% in medium lighting and 86% in low-lighting with a processing time of 0.5 s/frame on an average. Adaptive brightness correction and pose normalization were used to enhance performance under changing environmental settings. The cost-effective and scalable solution minimizes proxy attendance and manual recording mistakes, thus suitable for educational rollout. Yet, there are limitations such as lower accuracy in extreme pose variation and occlusions. Convolutional neural networks will be investigated in future research to increase robustness and recognition reliability.