Advanced Face Detection Using SVM and Haar-like Features: A Comprehensive Approach in AI and Image Processing

Face detection is a vibrant research field spanning artificial intelligence, image processing, video surveillance, human-machine interaction, identity authentication, machine learning, pattern recognition, and so on. While humans effortlessly recognize faces, computers face challenges due to facial variability. The complexity lies in factors like expressions, lighting changes, and head positions. Thus, solving the problem of face detection has an important academic value and provides important enlightenment. There are many aspects to face detection such as facial expression, head position, changing lighting conditions and many others. To deal with face detection problems various approaches are used such as template matching methods, knowledge-based methods, feature invariant methods, and appearance-based methods. The proposed approach for face detection utilizes Support Vector Machine (SVM) and Haar-like Features. It begins by generating an integral image from a binary image measuring 19 x 19 pixels. Following this, the Haar feature extraction method is employed to obtain Haar values from both face and non-face isommages. A set of 200 face and non-face images from MIT database have been used. Features have been reduced five times to make different features sets of images. These features are passed to Support Vector Machine as input parameters and processed to classify face and non-face datasets. Support Vector Machine is used to create, train and test the neural network. Features are extracted on MATLAB and further processed in SVM using NeuroSolutions tool. The results are evaluated based on two scenarios as varying number of images and features.