Pose and occlusion invariant face recognition system for video surveillance using extensive feature set

Abstract

Face recognition presents a challenging problem in the field of image analysis and computer vision. Different video sequences of the same subject may contain variations in resolution, illumination, pose, and facial expressions. These variations contribute to the challenges in designing an effective video-based face-recognition algorithm. In this proposed method, we are presenting a face recognition method from video sequence with various pose and occlusion. Initially, shot segmentation process is done to separate the video sequence into frames. Then, face part is detected from each frame for further processing. Face detection is the first stage of a face recognition system. After detecting the face exactly the facial features are extracted. Here SURF features, appearance features, and holo-entropy is used to find out the uniqueness of the face image. The active appearance model (AAM) can be used to find out the appearance-based features in the face image. These features are used to select the optimal key frame in the video sequence which is based on the supervised learning method, modified artificial neural network (MANN) using bat algorithm. Here bat algorithm is used for optimising the weights of neurons. Finally, based on the feature library, the face image can be recognised.