Hair Color Classification in Face Recognition using Machine Learning Algorithms
Saman Sarraf · American Scientific Research Journal for Engineering, Technology, and Sciences (Global Society of Scientific Research and Researchers) · 2016
Security through automatic human identification is critically important today, and this is largely due to the high volume of communications. Most methods used to identify individuals often use biometrics information, such as facial characteristics. Therefore, face recognition and classification have garnered great interest among computer vision researchers over the past decade. This pattern recognition problem is divided into several subcategories, such as eye or hair detection and classification. Hair is a salient feature in the human face and is one of the most important cues in face detection and recognition. Accurate detection and presentation of the hair region is one of the key components in the automatic synthesis of human facial caricature. In this work, hair color classification through feature extraction and machine learning methods was performed. The impacts of different features and classifiers were investigated using color samples. Support vector machines (SVM) and Kth nearest neighbors (K-NN) were trained by variety sets of statistical and color features, and the trained models were validated. Additionally, the effects of the size of datasets and feature dimensionality reduction were obtained. The best accuracy rate of 99% was achieved through a support vector machine with radial basis kernel function (SVM-RBF) using nine selected statistical and color features.