The invention discloses a human face recognition method based on a multi-camera model. The method comprises the steps of 1 first multi-camera information acquisition, 2 human face detection and training, 3 feature storage, 4 human face (structural data mode), 5 secondary multi-camera information acquisition, 6 database-related information acquisition and 7 double-layer heterogeneous depth neural network algorithm comparison. According to the invention, a secondary multi-camera information acquisition technology is used; a double-layer heterogeneous depth neural network algorithm comparison technology is used; the problem that an identity card photo stored in an identity card chip has a small pixel in a human face area, is blurred, is generally a photo of years before, and greatly differs from a photo collected in the field is solved; the problem that a human face has many shelters, such as beards, glasses, hats and the like, is solved; the problem that the environmental light intensity and the observation angle of the human face are changed is solved; the problem that the human face is difficultly recognized is solved; and the provided human face recognition method is economic, real-time and accurate.