Deep learning method for automatically describing landslide hidden danger according to ground penetrating radar image

The invention discloses a deep learning method for automatically describing a landslide hidden danger according to a ground penetrating radar image, and belongs to the technical field of image description. The method comprises the following steps: S1, generating an interpretive text, and constructing an image data set of landslide hidden danger development; s2, constructing a model and outputting a text; and S3, generating an interpretive text result. According to the method, an automatic deep learning model framework is provided, the development state of the landslide hidden danger image is determined, the explanatory text is generated, and an algorithm for generating a plurality of explanatory texts from a single image is provided, so that identification and comprehensive interpretation of the landslide hidden danger development form are realized; therefore, the application of image description in landslide hidden danger investigation is promoted, and the labor cost consumed by ground penetrating radar anal