Remote sensing image fusion method and system based on multi-scale dynamic convolutional neural network

Abstract

The invention discloses a remote sensing image fusion method and system based on a multi-scale dynamic convolution neural network, and the method comprises the steps: firstly enabling a high-resolution panchromatic image and a low-resolution multispectral image to dynamically generate a multi-scale filter through a multi-scale filter generation network, and then carrying out the multi-scale dynamic convolution of the filter and the panchromatic image; and properly weighting detail features obtained by dynamic convolution by using a weight generation network, enabling the weighted multi-scale detail features to pass through two convolution layers to obtain a final detail image, and adding the detail image and the low-resolution multispectral image to obtain a fused image. According to the invention, multi-scale local adaptive dynamic convolution is adopted, a local adaptive filter can be dynamically generated at each pixel position according to each input image, the adaptability of thenetwork is enhanced, the generalization ability of the network is improved, a good fusion effect is obtained, and the method and system can be used for target detection, target recognition and the like.

Description

technical field [0001] The invention belongs to the field of image processing, and in particular relates to a remote sensing image fusion method and system based on a multi-scale dynamic convolutional neural network. Background technique [0002] In order to efficiently utilize and integrate images from different sensors, image fusion technology has been developed rapidly and has been widely used in civil, military, medical, computer vision and other fields. In image fusion technology, the fusion of multispectral image and panchromatic image is the most widely used. Multispectral images contain rich spectral information but low spatial resolution; while panchromatic images have rich spatial details but lack spectral information, the spatial resolution of multispectral images can be improved through the fusion of multispectral images and panchromatic images, making It is better used in the identification and detection of surface objects and other applications. [0003] Ther...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the traditional remote sensing image fusion method based on convolutional neural network, the filter is fixed after training, and the adaptability is poor, resulting in the loss of fusion image details and spectral distortion, and provides a multi-scale dynamic convolutional neural network. Network remote sensing image fusion method and system

Images