Different-spectral-coverage infrared image transformation method based on gradient restriction generative adversarial network

Abstract

The invention discloses a different-spectral-coverage infrared image transformation method based on a gradient restriction generative adversarial network. The method utilizes a generator network to transform an infrared image to be detected is a different-spectral-coverage infrared image. The training of the generator network comprises the following steps that: dividing a plurality of paired infrared image pairs of two types of spectral coverage into a training sample pair and a test sample pair, and constructing the generative adversarial network, wherein the generative adversarial network comprises the generator network and a discriminator network; inputting the training sample into the generative adversarial network, calculating the losses of the generator network and the discriminatornetwork on the basis of the loss function of a gradient restriction, and carrying out back propagation to update the generator network and the discriminator network; and dividing the test sample pairinto a plurality of test image blocks, utilizing the test image blocks to test the generator network, obtaining a transformation picture, and obtaining the trained generator network when the mean value of a local relevant value between the transformation picture and the test image blocks is greater than or equal to a threshold value, obtaining the trained generator network. An image transformationeffect is good.

Description

technical field [0001] The invention belongs to the field of infrared image processing, and more specifically relates to a heterospectral infrared image transformation method based on a gradient-constrained generation confrontation network. Background technique [0002] With the advancement of science and technology and the continuous development of camouflage and jamming technologies, single-band broadband target detection and recognition can no longer meet the needs of target recognition under complex conditions. Multi-band detection and multi-feature information fusion technology will soon become an important means of target recognition. Infrared multispectral images can provide not only the spatial information of the target, but also the spectral information of the target, which provides the data basis for the analysis of infrared target feature extraction, fusion and recognition. [0003] The rapid development of infrared imaging technology and its great application val...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above defects or improvement needs of the prior art, the present invention provides a heterospectral infrared image transformation method based on a gradient-constrained generative adversarial network, thereby solving the problem of high cost, time-consuming, and difficulty in obtaining an approximate real scene in the prior art. Technical Issues of Multispectral Simulation Images

Images