This monograph presents the basic concepts, methods, and models of multispectral image recognition based on oscillatory and diagonalized neural networks in a logical sequence. In addition, the monograph contains sections on the modified Hecht-Nielsen theorem, optimizing the size of a neural network taking into account the nonlinear generalized error of the neural network model, creating a pseudo-spin neural network with reduced learning time, and creating a decision support model based on a diagonalized neural network.