Multiple Face Recognition Using Self-adaptive Differential Evolution and ORB

Abstract

Face recognition (FR) applications have been intensively studied in the areas of Computer Vision specially with the wide use of bio-metrics as a method of security access. Hence, working on still images with multiple faces (SIMF) is still a challenge due to the diversity of features that may be present in the images and different conditions from which the images are usually captured. In addition, FR algorithms may not perform well under uncontrolled conditions due to the illumination variation and other related image conditions. Hence robust approachesshould be developed to perform well under complex conditions. Based on this context, in the present work, we proposed a new approach for recognizing faces in SIMF obtained under different conditions using an optimization algorithm with feature extraction method based on interest points (IP). Our algorithm is constructed using the Self Adaptive Differential Evolution (SaDE) algorithm and ORB, well-known key point detector and descriptor. A specific base of SIMF obtained under different illumination conditions, rotation, scale, and occlusion is used to validate the algorithm. The recognition process is conducted with the aid of optimization algorithm searching for a possible best match of a template face image (TFI) in the SIMF. During the search process, numerous target cut images (TCI) from SIMF are evaluated using descriptor distance measure. According to the experiment results, it can be observed the present approach is suitable for real-world FR applications with complex conditions.