论文部分内容阅读
Recently,the emerging concept of smart city is attracting more and more focus.In this paper,we present a novel depth enhanced face recognition system for smart city.We propose the Extended Local Line Derivative Pattern (ELLDP) which extends the Local Derivative Pattern (LLDP) to an arbitrary angle.We prove that the direction of gradient is the optimal angle.To make the angle fixed for the gallery and test faces of the same person,we calculate the gradient direction of the depth map to estimate the gradient direction of the intensity image.In the system the intensity image and depth map of a face are divided into blocks.The accumulated gradient orientation histogram of the depth block is used to find the optimal angle of the ELLDP for the intensity image block.A histogram is obtained for each block.The histograms are concatenated to one as the final feature representation.At last,the probe face is assigned the identity with the smallest classification distance.Extensive experiments are conducted on 3 sub-databases from CurtinFaces database which contain variations in illumination,expression,pose and disguise.The results that our approach achieves the highest recognition rate and much gentler ROC curve compared with other methods demonstrate the robustness and superiority of our approach.