Unmanned aerial vehicle night image processing method

The invention belongs to the technical field of image processing, and particularly relates to an unmanned aerial vehicle night image processing method, which comprises the following steps of: acquiring a night image by using an unmanned aerial vehicle, and acquiring a non-night image at the same position by using the unmanned aerial vehicle; processing the collected night images and non-night images, and matching and numbering the processed images to obtain images matched in pairs; and constructing a multi-stage cross-layer cascade neural network model, performing supervised training on the model by using the matched paired images, adjusting parameters of the model according to a loss function, and using the model after parameter adjustment to process the unmanned aerial vehicle night image. The night image collected by the unmanned aerial vehicle is recovered through a multi-stage progressive processing mode, the model structure is simple, and the image recovery effect is good. 本发明属于图像处理技术领域，具体涉及一种无人机夜间图像处理方