Remote sensing data scale adaptive adjustment method and device, storage medium and equipment

Abstract

The invention relates to a remote sensing data scale adaptive adjustment method and device, a storage medium and equipment, and the method comprises the steps: carrying out the ground object segmentation of a remote sensing image of a target region based on a mean shift algorithm, constructing a minimum bounding rectangle of each ground object, determining a plurality of image scale combinations according to the length values of the longest sides of a plurality of minimum bounding rectangles, and obtaining a plurality of image scale combinations; based on a grid search method and a pre-trained remote sensing classification model, an optimal image scale combination and an optimal scale sampling interval with the highest remote sensing classification precision in an image set are obtained, and adaptive adjustment of model input image scales is realized. The pre-trained remote sensing classification model is utilized to perform remote sensing classification on the remote sensing images with the optimal image scale combination and the optimal scale sampling interval, so that local optimum caused by manual image scale setting is avoided, and the remote sensing classification accuracy is improved.

Description

technical field [0001] The invention relates to the field of remote sensing images, in particular to a remote sensing data scale adaptive adjustment method, device, storage medium and equipment. Background technique [0002] In the prior art, a Convolutional Neural Network (CNN) is usually used to extract features contained in remote sensing images. The convolutional neural network usually includes structures such as an input layer, a convolutional layer, a pooling layer, and an output layer. It automatically extracts the feature features contained in the remote sensing image through the filter in the convolutional layer. The input of the convolutional layer The data comes from the remote sensing image input by the input layer, and the scale of the remote sensing image directly affects the size of the receptive field of the convolutional layer. [0003] At present, remote sensing images are usually sampled in a fixed image scale interval manually set to improve the accuracy...

AI Technical Summary

Problems solved by technology

[0003] At present, remote sensing images are usually sampled in a fixed image scale interval manually set to improve the accuracy of remote sensing classification. However, the remote sensing images obtained by the above sampling method have a large randomness, which affects the accuracy of remote sensing classification.

Images