A radar image de-noising method based on GAN

Abstract

The invention relates to a radar image de-noising method based on a GAN (Generative Adversarial Network). The method includes: the establishment of a radar simulation data set; the addition of the noise in a radar time frequency image; the establishment of the GAN, wherein the GAN is comprised of a generative network and discrimination network, a convolution nerve network structure is adopted by the two networks, a convolution nerve network is established as the generative network based on a residual error network, the front two convolution layers are designed as a convolution kernel with a step size of 2 so that the down-sampling is realized, the operand is reduced and the abstractness feature is extracted, the unit that keep the scale of the feature image in the residual error network isfollowed, and the pooling layer in the network is replaced by the convolution layer with invariant scale of feature image; a fully connected layer is deleted which is located after the convolution layer and used for the classification of the image, the up-sampling for the image is realized by utilizing two deconvolution layers, the output feature of the shallow convolution layers and the output feature of the deep convolution layers are added and further processed, the reuse of the feature is realized. The radar image de-noising method also includes training of the GAN.

Description

technical field [0001] The invention belongs to the field of radar image processing, signal processing and deep learning, and relates to related applications such as human body detection and tracking based on radar images. Background technique [0002] Micro-Doppler radar has always been widely used in the military field, and plays a vital role in military detection, anti-terrorism operations and security activities. With the rapid development of signal processing and radar-related applications, the application field of micro-Doppler radar is gradually inclined to civilian use. Micro-Doppler radar actively emits electromagnetic wave signals based on the Doppler principle. When the detection area contains moving objects, the signal emitted by the radar will be modulated to generate the Doppler effect to reflect the echo carrying the target movement information. Signal. When the detection object is not a rigid body, but a complex movement involving multiple parts, the micro-...

AI Technical Summary

Problems solved by technology

The invention utilizes the generative confrontation network in the deep learning algorithm to realize the denoising of the radar time-frequency image in the environment of low signal-to-noise ratio, and aims at the difficult condition of radar data acquisition, realizes the training of the generative confrontation network by simulating the radar image, and improves the radar When the image noise is serious and the quality is low

Images