Land utilization classification and change prediction method based on deep learning

Abstract

The invention provides a land utilization classification and change prediction method based on deep learning, and the method comprises the steps: making land coverage type training samples and corresponding land utilization type training samples, constructing a deep learning semantic segmentation network model, constructing a type conversion network model based on a gating mechanism, and carrying out the training of the model conversion network model, generating a land utilization classification model according to the deep learning semantic segmentation network model and the type conversion network model, obtaining a land utilization classification image according to the land utilization classification model, selecting driving factors to construct a prediction model, and analyzing and concluding land utilization spatial-temporal change characteristics and laws in different periods according to the land utilization classification image, and using the prediction model to predict the future land utilization change. According to the land utilization classification and change prediction method based on deep learning, land utilization classification and change prediction can be accurately realized, and a basis can be provided for a series of work such as land utilization dynamic change prediction.

Description

technical field [0001] The invention relates to the technical field of remote sensing image information, in particular to a land use classification and change prediction method based on deep learning. Background technique [0002] Land use classification technology based on high-resolution remote sensing images has been widely used in large-scale land use surveys, but the existing land use classification technology still has difficulties in extracting classification features of land objects, many noise interference factors, and insufficient classification results. In addition, the types of land use are diverse and the basis for classification is complex, and some categories are composed of a variety of different features, resulting in a complex internal structure. Generally, classification methods that rely on remote sensing image features cannot accurately classify complex land use types. . The existing land use classification and change prediction are not combined. When p...

AI Technical Summary

Problems solved by technology

[0002] Land use classification technology based on high-resolution remote sensing images has been widely used in large-scale land use surveys, but the existing land use classification technology still has difficulties in extracting classification features of land objects, many noise interference factors, and insufficient classification results. In addition, the types of land use are diverse and the basis for classification is complex, and some categories are composed of a variety of different features, resulting in a complex internal structure. Generally, classification methods that rely on remote sensing image features cannot accurately classify complex land use types.

The existing land use classification and change prediction are not combined. When predicting land use change, ordinary classification techniques are often used for classification, resulting in low prediction accuracy, and it is impossible to carry out land use dynamic change prediction, crop yield prediction, natural disasters, etc. A series of work such as prevention and control and rational organization of land use

Images