Pose-Invariant Face Recognition in Videos for Human-Machine Interaction

摘要: Human-machine interaction is a hot topic nowadays in the communities of computer vision and robotics. In this context, face recognition algorithms (used as primary cue for person's identity assessment) work well under controlled conditions but degrade significantly when tested real-world environments. This mostly due to difficulty simultaneously handling variations illumination, pose, occlusions. paper, we propose novel approach robust pose-invariant human-robot based on real-time fitting 3D deformable model input images taken from video sequences. More concrete, our generates rectified image irrespective with actual head-pose orientation. Experimental results performed Honda database, using several manifold learning techniques, show distinct advantage proposed method over standard 2D appearance-based snapshot approach.