Heterogeneous feature fusion-based optimal face image acquisition in visual sensor network

High quality face image acquisition from huge video data obtained in visual sensor network is of great significance in applications related to face processing, such as face recognition and reconstruction. This paper proposes an optimal face image acquisition method in visual sensor network, which is based on collaborative face frames acquisition and heterogeneous feature fusion-based face quality assessment. Gaussian-probability-distribution-based multi-view data fusion and kalman filter are used for collaborative target localization and tracking. To achieve primary screening of face frames, a lightweight face frames quality evaluation method is presented. Importantly, new face quality assessment criterion calculation methods are proposed to make fine screening of face images more applicable in visual sensor network. The new face quality assessment criterion calculation methods are based on heterogeneous feature fusion of pedestrian tracking and static face image features analysis. Fuzzy inference engine is used to combine these criteria to generate a face quality assessment score. Experimental results show that the proposed method can acquire optimal face images accurately and robustly.