Object-oriented graph neural network unsupervised remote sensing image change detection method

The invention discloses an object-oriented graph neural network unsupervised remote sensing image change detection method, and relates to the field of remote sensing detection. The method is used for improving the high-resolution remote sensing image change detection precision and the automation level, and comprises the steps of performing image segmentation and feature selection on two time phases; calculating a multi-level weighted difference degree of the two time phase image objects, obtaining an optimal segmentation threshold value by using an EM algorithm and a Bayesian minimum error equation, and preliminarily dividing all the image objects into two change categories of change objects and non-change objects by using the threshold value; adaptively screening a certain proportion of image objects which are most likely to change and not change as training samples; a jump connection structure is introduced, the feature fusion radius is adaptively selected, the number of network layers is increased, and hig