Semi-visible face detection for safety in unconstrained crowd environment

Abstract

Face recognition has gained interest due to largely growing public concerns for safety, specifically due to many terror activities around the world. These problems of terror increased the demand of suitable face recognition system in security applications. Face detection is the initial step for face recognition and it has been widely researched and examined. Numerous methods have been proposed with high detection rates in restricted scene. However, in unrestricted scene those method failed to perform accurate detection. The main challenges to deal in uncontrolled scene are detection of multiview, illuminated, occluded and blur faces. This research paper has studied the issues of face detection in unconstrained scene. More specifically, it has discussed the existing occluded face detection methods. Based upon the survey of different face detection methods, a half occluded face detection method is proposed. In order to deal with half occluded faces, new type of image features are proposed.