Research on the recognition technology of human relationship in face image based on deep learning

Face images contain rich and complex information, and similar genes are directly expressed as facial resemblance. In this paper, based on the theoretical background of biology and aiming at the wider application of artificial intelligence in the field of face image analysis, deep learning based recognition of the kinship of people in face images is studied in depth. Kinship recognition is a new and challenging research. In order to recognize the kinship between characters from face images, this paper defines a new correlation feature to represent the correlation between two characters. Based on the idea of multi-layer self-encoder to input and reconstruct the target, and combined with the powerful feature extraction ability of deep convolutional neural network, this paper designs a new deep convolutional self-coding neural network, and the correlation feature is the activation value of the deep hidden layer of this deep neural network. The deep convolutional self-coding neural network USES back propagation algorithm in a supervised way. With the deepening of the network hierarchy, the identity features of representative characters are extracted from face images step by step, and the identity features of a pair of characters are fused into the associated features that represent the relationship between them. Based on deep learning, the associated features extracted from face images can effectively identify a given character relationship.