SAR image automatic colorization method based on GANIlla

Abstract

The invention discloses an SAR image automatic colorization method based on GANIlla, and relates to satellite image processing. The method comprises steps of data set pairing: randomly generating sites on a map, taking a selected site expansion matrix as a selection area, respectively extracting a VV wave band in Sentinel-1 and B4, B3 and B2 wave bands of Sentinel-2 as RGB values, and visualizing data in a JSON format based on python to obtain a color migration data set; dividing the obtained data into two types, namely data of a land area and data of an ocean area; increasing the data of the land area through rotation, local amplification and overturning; carrying out wiener filtering and random color adding respectively on the data of the ocean area; and carrying out training on the Sentinel data after the data is increased by using a GANIlla network. The complex work of establishing a network is avoided, the data pairing speed is improved, the data acquisition difficulty is reduced, and the remote sensing image quality is greatly improved.

Description

technical field [0001] The invention relates to the field of satellite image processing, in particular to a GANilla-based SAR image automatic colorization method. Background technique [0002] The Synthetic Aperture Radar (SAR) images based on the active microwave imaging mechanism captured by the Haisi-1 satellite have the observation characteristics of all-day, all-weather, and not affected by severe weather. SAR images have grayscale and texture characteristics, and the SAR image textures presented by different ground objects are different. Therefore, SAR images can be used to realize target detection based on the differences in grayscale and texture features between ground objects. In addition, SAR satellites have good observation penetration performance, can penetrate clouds, discover hidden target information, distinguish false targets, and have certain display capabilities for dynamic targets, and have unique advantages in space-to-earth observation and military recon...

AI Technical Summary

Problems solved by technology

The current image matching often relies on selecting well-imaged feature images for candidates, and establishing the correspondence between sites according to the imaging distribution, which requires a lot of manual intervention

In 2020, Lloyd et al. proposed to use the framework of three neural networks, Goodness Network, Multi-scale Matching Network, and Outlier Reduction Network, to perform data fusion comparison and improve The pairing rate of the dataset, but the quality of the final correspondence set is still limited by the goodness network and the outlier reduction network

Images