Performance Evaluation of Identity Based Face Recognition Models in IoT Based Smart Home

In recent years, the field of image recognition has been significantly advanced by deep learning, which offers highly effective techniques for feature extraction and classification. Face detection and recognition is a fundamental computer vision technology that enables systems to automatically identify and locate human faces in digital images or video frames. It supports a wide range of applications, including surveillance, facial recognition, and emotion analysis. OpenCV, a widely used open source computer vision library, provides an extensive suite of tools and pre-trained models that streamline the implementation of these technologies. One of the most popular algorithms within this framework is Local Binary Patterns Histograms (LBPH), known for its resilience to varying lighting conditions and its consistent performance across diverse environments. Its low computational demand also makes it an ideal choice for real-time processing in resource-limited settings. These features make LBPH a practical alternative to more resource intensive deep learning approaches in certain use cases. In this paper, we investigated the effectiveness of the LBPH algorithm in security, face identity and authentication applications and compares it with other face recognition methods including Fisherfaces, Eigenfaces and deep neural network based techniques. All algorithms were implemented and evaluated across multiple datasets, with performance measured in terms of accuracy, reliability, and real-world applicability. The analysis also incorporates key evaluation metrics and Receiver Operating Characteristic (ROC) curves for comprehensive performance assessment.