Face Recognition Based Person Specific Identification for Video Surveillance Applications

Abstract

Face detection is an important aspect for applications like biometrics, video surveillance and human computer interaction. Videos provide abundant information and also that can be leveraged by temporal variations in pose, expression changes and occlusion. These challenging problems motivate to identify the specific person by face recognition for video surveillance applications. This paper presents face detection and recognition algorithm to identify/recognize the wanted person in a surveillance video. First, face detection is done by 'Viola-Jones' algorithm. Subsequently, the detected face has been cropped to recognize the specific person's face. For the cropped faces, clustering is applied to cluster the face parts. For the detected face parts HOG and LBP features are obtained. Existing approaches use Histogram of Oriented Gradients (HOG) and Local Binary Pattern (LBP) separately to recognize face in static images. The contribution of the work is to implement HOG and LBP for surveillance video and combine both the features to address the issues such as pose variations, illumination changes, expression changes and occlusion for face recognition. SVM classifier is used to classify the weak and strong features and strong features are used to recognize the person. The proposed algorithm has been tested on various datasets and its performance is found to be good in most cases. Experimental results show that the method of detection and recognition achieves very encouraging results with good accuracy and simple computations.