Face recognition algorithm based on cascading BGP feature fusion

BGP feature is a simple and efficient face recognition descriptor which is robust to light, expression and occlusion. Normally, the feature extraction of BGP is performed once on the face, and the texture information obtained is not rich enough. In this paper, a multi-level cascaded BGP extraction method is applied to obtain richer facial texture information. Firstly, feature extraction is performed on the original face using BGP. Then, feature extraction is performed on the obtained face feature images in the same way to obtain multi-level concatenated features. Finally, the histograms of the feature images are spliced and combined on the face features. The algorithm is fully validated in Yale and extended YaleB, and the recognition accuracy is significantly improved compared with the original BGP algorithm.