Method for estimating DOA (direction of arrival) of meter wave radar based on two-dimensional convolution neural network

Abstract

The invention belongs to the technical field of radar, and discloses a method for estimating the DOA (direction of arrival) of a meter wave radar based on a two-dimensional convolution neural network.The method comprises the steps of obtaining P plots as a training set; calculating the covariance matrix of each plot in the training set and obtaining the corresponding phase average matrix and phase standard deviation matrix, wherein phases corresponding to upper triangular elements form an upper triangular element phase matrix; using the phase matrix corresponding to the i plot and subjected to zero filling and rearrangement as the input of the convolution neural network, and the output matrix of the convolution neural network corresponding to the i plot is obtained; modifying thenetwork parameters of the convolution neural network according to an objective function; and obtaining an actually measured target plot, and inputting the phase matrix of the actually measured targetplot into the convolution neural network to reconstruct the covariance matrix of the actually measured target plot to estimate the DOA of the target plot. Thus the DOA estimation problem is transformed into a pure regression problem.

Description

technical field [0001] The invention belongs to the technical field of radar, and in particular relates to a method for estimating the DOA of a meter-wave radar based on a two-dimensional convolutional neural network, which can be used for estimating the angle of arrival (DOA) of a meter-wave radar in a low-elevation angle and multipath environment. Background technique [0002] At present, in order to meet their strategic needs, most stealth aircraft or fighters use low-altitude / ultra-low-altitude flight methods to effectively attack their strategic targets. The meter wave radar has a longer wavelength, which has a better anti-stealth effect than other higher frequency bands. However, due to the wide beam, when the target is flying at low altitude / ultra-low altitude, there is a serious beam "hitting the ground" phenomenon, and the multipath signal reflected by the ground is received by the radar, which greatly weakens the power of the radar and its measurement accuracy. ...

AI Technical Summary

Problems solved by technology

[0005] In view of the above problems, the object of the present invention is to provide a DOA estimation method for meter-wave radar based on two-dimensional convolutional neural network, which can not only effectively solve the problems of model mismatch and insufficient prior information in meter-wave radar engineering practice, but also There are no disadvantages of existing research results, and the DOA estimation problem is completely transformed into a pure regression problem

Images