Image Analysis for Face Recognition and Detection Using Machine Learning Algorithms

Dominant methods for security purposes have emerged as a public safety concern for modern society. In this study, we have have incorporated different algorithms like Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), Principal Component Analysis (PCA), and Voila-Jones (VJ) for the task of face recognition and detection. Each of these applications has its own distinct advantages and disadvantages. For example, PCA is often regarded as a simpler and more appropriate approach for differentiating faces within a given image. On the other hand, the Voila-Jones algorithm excels in the face detection mechanism, offering superior capabilities for detecting faces in two-dimensional settings. While these applications have demonstrated successful results, they may encounter challenges in accurately representing faces due to various factors. In this study, we have made a comparative investigation of the performance, accuracy, and effectiveness of these algorithms in their application in face recognition and detection tasks. This study was conducted and tested based on well-known face datasets, such as ORL and Yale data sets. We divided these datasets into training and testing sets to prepare the data. We then proceeded to perform face recognition and classification, considering the selection of input data. Finally, we computed the system’s accuracy. Moreover, this study focused on conducting face recognition and detection in 2-Dimenitional settings. For performance evaluation, we used various testing and training images. In the testing phase with 45 images, PCA showed an accuracy of 97.78%, whereas Voila-Jones algorithm achieved a detection accuracy of 100%. Furthermore, when testing with 40 images, LDA achieved an accuracy of 100%, however, KNN demonstrated an accuracy of 98.5%. However, the Viola-Jones algorithm achieved a slightly lower detection accuracy of 97.5%. Overall, our results showed that LDA outperformed the other methods and demonstrated better results.