Face recognition analysis by developing feature operators in virtual reality

The face feature points are required to be found automatically to detect human emotion from face images, to detect age by analyzing human face from face image, to reconstruct 3D face model by using multi - human face images, develop augmented reality applications on human face. In the present thesis, an algorithm and software have been developed to detect human face and find interest regions and interes points in face from human face images taken from different angles and distances under standardized conditions automatically. The data used for test purposes is obtained from a face database which is consist of face images taken under a 1000W halogen lamp by a 3D structural-illuminated 3D digitizer device with a resolution of 1600x1200. Some cases are excluded from this study, such as some parts of the face is to be covered by other objects, beard, mustache, facial expressions and gestures. The application developed in presented thesis consists of three sequential steps. Search region is made more narrowed after every step. Algorithms detect faces, search interest regions and facial feature points. As a result of tests carried out on 360 different face images with 1100 * 1400 pixel sizes, which belongs to 35 different people, 34 of them are women and 11 of them are men, it is achieved 100% face detection accuracy; 2,04086 pixel for facial feature points detection and 1,83971 accuracy for 3D point cloud generation. Conjugate images are being matched easily by means of smart feature regions and matching feature points.After stereo matching of images, bundle adjustment algorithm is run. Applying the epipolar geometry rules on the new balanced points, three-dimensional reconstruction algorithms have been run and three-dimensional point cloud face is generated.The data obtained is used in an augmented reality application. Augmented reality application recognizes faces which are defined in application database, creates a marker by using facial feature points automatically and add 2D / 3D virtual objects on real world.Key words: Facial feature points, feature operators, face detection, face recognition, augmented reality