OCCLUDED FACE DETECTION IN UNCONSTRAINED CROWD SCENE

Abstract

Face verification system’s gain significant attention in last few years due to increase security concern in public and private places. Face detection is the most important and initial stage in automatic face verification system. It helps to determine the existence of faces in an image and return the position and location ofthe face. Face verification system accuracy depends on face detection. Human face therefore, face detection is challenging in is not always frontal and has many variations, unconstrained scenarios. One main challenge of face detection is occlusion. In existing research work, a method is proposed to address this challenge in unconstrained face detection. Proposed work introduces an occluded face detection system. The Viola-Jones algorithm which is the frontal face detector has been adopted in the . Two cascade proposed approach along with new type of features called free rectangular features classifiers are used in which one is trained for holistic faces and second is trained on half occluded faces, both of the classifiers are used in parallel to work in unconfined scene. Additionally, for improvement the correctness and adeptness of the system, skin color models are applied for improvement of detection results and removing offalse positive detection. Three experiments had been carried out to test the effectiveness ofthe system. First experiment is performed on FDDB public database, second experiment on the unconstrained crowd images collected through internet and third experiment on uncrowded images. Quantitative and qualitative evaluation of results has been performed. The experimental results showed that the is effective in detecting the half occluded and holistic faces in unconstrained scene.