Physical driven neural network-based color generalization imaging method of through scattering medium

The invention discloses a physical driven neural network-based color generalization imaging method of a through scattering medium. The method comprises the following steps of: establishing a neural network in which a physical model and a neural network model are driven by each other, and establishing system configuration for acquiring experimental image data, and preprocessing collected image data and sending the preprocessed image data to the neural network so as to perform image restoration and reconstruction. According to the method of the invention, on the basis of effective combination of speckle correlation and speckle redundancy physical prior and strong data mining and mapping capabilities of the multi-channel convolutional neural network, high-quality recovery of a target can be realized under the condition that a multispectral complex target penetrating through an unknown scattering medium can be realized only by using speckle data of one medium; and the generalization imaging of a physical percepti