Semi-supervised learning deep valley intelligent identification method based on remote sensing image

The invention discloses a semi-supervised learning deep valley intelligent identification method based on a remote sensing image, and belongs to the technical field of image processing, and the method comprises the steps: S1, selecting a satellite remote sensing image covering a deep valley landform in a multi-scene manner and digital elevation model (DEM) data, and building a deep valley landform intelligent identification data set; s2, constructing a deep convolutional neural network coding model by using a ResNet50 algorithm structure, a channel and a space attention mechanism based on a deep learning coding principle; s3, constructing a deep river valley intelligent identification main decoder and an auxiliary decoder model under a characteristic disturbance function based on a semi-supervised learning and decoding theory; and S4, constructing and training a deep river valley intelligent identification model, and carrying out intelligent identification on the deep river valley by utilizing the deep river