CN-122017748-A - Radar interference suppression and target identification integrated method and system

CN122017748ACN 122017748 ACN122017748 ACN 122017748ACN-122017748-A

Abstract

The invention discloses an integrated radar interference suppression and target identification method and system, which are used for constructing an end-to-end complex value neural network for radar target identification under radio frequency interference, training the network by using a complex value radar training data set containing interference-free and interference-containing complex value radar image data and target categories as supervision labels, and simultaneously learning anti-interference and target identification, wherein the network structure comprises an anti-interference and target characteristic enhancement module and a target identification module, the anti-interference and target characteristic enhancement module is used for separating targets and radio frequency interference from input complex value radar images, suppressing the radio frequency interference and highlighting the characteristics of the targets, and the target identification module is used for learning and extracting the characteristics of the targets from the output of the anti-interference and target characteristic enhancement module. The method and the device can solve the problem that in the prior art, the identification performance is reduced due to error accumulation in the cascade processing flow, and are suitable for radar image target identification tasks under radio frequency suppression interference.

Inventors

HUANG YAN
LI ZEHAO
MAO YUAN
ZHANG HUI
YU XUTAO

Assignees

东南大学

Dates

Publication Date: 20260512
Application Date: 20260119

Claims (10)

1. The radar interference suppression and target identification integrated method is characterized by comprising the following steps of: Acquiring interference-free complex-valued radar echo data, acquiring or manufacturing an interference-containing complex-valued radar data set through simulation, and reserving the interference-free complex-valued radar image data as a supervision tag during training; The system comprises a training data set, a target identification module, an anti-interference and target feature enhancement module, a target identification module and a target identification module, wherein the training data set is used for training a network, and interference-free complex-valued radar image data and target categories are used as supervision labels, so that the network can perform anti-interference and target identification learning at the same time to obtain a trained end-to-end complex-valued neural network; Inputting the complex-valued radar image data to be detected into a trained end-to-end complex-valued neural network to obtain the identification result of the radar target under radio frequency interference.
2. The integrated radar interference suppression and target identification method of claim 1, wherein the creating of the data set includes: Obtaining interference-free complex-valued radar echo data Acquiring or obtaining radio frequency interference data with different intensities and different categories through simulation; adding radio frequency interference data to raw interference-free radar echo data along the fast time dimension Obtaining radar echo data containing interference ; The obtained radar echo data containing interference Imaging processing is carried out to obtain complex-valued radar image data containing interference At the same time, radar echo data without interference to the original Imaging processing is carried out to obtain interference-free complex-valued radar image data ; For complex-valued radar image data containing interference and no interference respectively And The same segmentation operation is carried out to obtain complex-valued radar image data containing single targets, up-sampling or down-sampling processing is carried out, samples are unified to the same size to obtain complex-valued radar target identification data sets containing interference and no interference, and corresponding target categories are recorded to be used as supervision labels.
3. The integrated radar interference suppression and target identification method according to claim 1, wherein: The anti-interference and target feature enhancement module is realized through a complex value U-Net network structure, and utilizes the encoding and decoding of the complex value U-Net network and a jump connection structure to realize the suppression of radio frequency interference in an input sample and the enhancement of target features; The target recognition module is realized through a complex-valued CNN network and a complex-valued full-connection layer, wherein the complex-valued CNN network is used for further extracting target characteristics, and then the complex-valued full-connection layer is matched with complex-valued modulo operation and softmax function to complete the task of recognizing the radar target.
4. The integrated radar interference suppression and target identification method according to claim 3, wherein the complex-valued U-Net network comprises a complex-valued convolution layer, a complex-valued ReLU activation function and a complex-valued maximization pooling layer operation; The complex value convolution layer has the following calculation formula: ; Wherein, the For the output of the complex-valued convolution layer, In the form of a complex-valued convolution kernel, For the input of the complex-valued convolution layer, A convolution operation that is complex valued, a convolution operation that is real valued, And The operation of taking a real part and an imaginary part respectively; the complex value ReLU activation function has the following calculation formula: ; The calculation formula of the complex value maximum pooling layer is as follows: ; ; Wherein, the Is complex value characteristic matrix In position The value of the position is taken out, For the complex-valued modulo operation, For the maximum pooling operation of real values, And (5) taking a position index corresponding to the maximum pooling operation value.
5. The integrated radar interference suppression and target identification method according to claim 4, wherein, in the complex-valued CNN network, the complex value batch normalization layer, the complex value average pooling layer and the complex value full-connection layer are also included in addition to the complex value convolution layer and the complex value ReLU activation function operation; the calculation formula of the complex value batch normalization layer is as follows: ; Wherein, the And Respectively a leachable scale transformation and translation transformation parameter, For inputting complex data of one Batch Is used for the average value of (a), For inputting complex data of one Batch For complex valued neural networks, The calculation formula of (2) is as follows: ; Wherein, the Calculating covariance; The calculation formula of the complex value average pooling layer is as follows: ; Wherein, the For a real-valued average pooling layer operation, A complex value characteristic matrix corresponding to the complex value pooling layer; The calculation formula of the complex value full-connection layer is as follows: ; Wherein, the For the output of the complex-valued fully-connected layer, Is a complex value weight matrix of a complex value full connection layer, For the input of the complex-valued fully-connected layer, Is biased with a complex value.
6. The radar interference suppression and target identification integrated method according to claim 1, wherein the training method of the end-to-end complex-valued neural network is as follows: The interference-containing complex-valued radar image data is used as input, the interference-free complex-valued radar image data is used for supervising the parameter learning of the anti-interference and target feature enhancement module, the target category supervises the anti-interference and target feature enhancement module and the target identification module at the same time, and the updating of network parameters is realized through a back propagation algorithm.
7. The integrated radar interference suppression and target identification method according to claim 6, wherein the training of the end-to-end complex-valued neural network adopts an improved loss function, the loss function is composed of a root mean square error loss term and a cross entropy error loss term, and the calculation formula is as follows: ; Wherein, the Super parameters for controlling the two proportions of the loss function; 、 the functions correspond to the anti-interference and target feature enhancement module and the target identification module respectively; representing input samples One hot code of the corresponding class c; Representing a complex valued radar image sample without interference, Representing complex valued radar images M, K and n represent the total number of training samples, the total number of data set sample classes and the total number of pixels of a single input sample, respectively; the probability of a sample class for the network output is calculated by a softmax function.
8. Integrated radar interference suppression and target identification system for implementing a method according to any of claims 1-7, comprising: The data set construction module is used for acquiring interference-free complex-valued radar echo data, acquiring or manufacturing an interference-containing complex-valued radar data set through simulation, and reserving the interference-free complex-valued radar image data as a supervision tag during training; The network construction and training module is used for constructing an end-to-end complex value neural network and carrying out radar target identification under radio frequency interference; training a network by using a training data set, and using non-interference complex-valued radar image data and a target class as supervision labels to enable the network to simultaneously learn anti-interference and target identification to obtain a trained end-to-end complex-valued neural network, wherein the end-to-end complex-valued neural network structure comprises an anti-interference and target feature enhancement module and a target identification module; The interference suppression and target identification module is used for inputting the complex-valued radar image data to be detected into the trained end-to-end complex-valued neural network to obtain the identification result of the target class.
9. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program when executed by the processor implements the steps of the radar interference suppression and target identification integrated method according to any one of claims 1-7.
10. A computer program product comprising a computer program, characterized in that the computer program when executed by a processor realizes the steps of the integrated radar interference suppression and target identification method according to any one of claims 1-7.

Description

Radar interference suppression and target identification integrated method and system Technical Field The invention relates to the technical field of radars, in particular to a radar interference suppression and target identification integrated method and system. Background The radar target recognition technology has wide application in the aspects of remote sensing detection, traffic control and the like. However, due to the increase of radiation sources, the radar is likely to be interfered by radio frequencies of the radiation sources such as a communication system, other radar systems and the like when receiving the echo, and the radio frequency interference signals can seriously destroy the target information in the radar echo, so that difficulty is brought to radar target identification. Therefore, how to efficiently implement radar target recognition under radio frequency interference is a problem that needs to be solved in the current radar target recognition technical field. At present, for radar target identification under radio frequency interference, a cascading processing flow is mainly adopted, namely anti-interference and radar target identification are carried out separately, radio frequency interference in radar images is restrained through an independent anti-interference module, and then the processed images are input into a target identification module to complete identification. While the conventional cascade processing flow has the advantages of low complexity, easy realization and the like, the performance is good in a weak interference scene, but the processing error of the anti-interference module is transmitted to a subsequent radar target recognition module to cause error accumulation and finally cause obvious reduction of recognition performance, and meanwhile, in the conventional radar target recognition technology, deep learning is mostly used as a core, radar images are complex images essentially and comprise amplitude and phase dual information, but the conventional method mostly adopts a neural network with real weight, only utilizes the amplitude information of the radar images to ignore target characteristics contained in the phase information, so that the utilization of the radar image information is insufficient, and the improvement of the recognition performance is further restricted. Disclosure of Invention The invention aims to provide an integrated method and system for radar interference suppression and target identification, which fully utilize amplitude and phase information in radar images, solve the problem of error accumulation in the traditional cascade processing flow, realize an end-to-end network structure integrating anti-interference and target identification tasks and improve the performance of radar target identification tasks under radio frequency interference. In order to achieve the above object, the first aspect of the present invention provides a radar interference suppression and target identification integrated method, comprising the following steps: Acquiring interference-free complex-valued radar echo data, acquiring or manufacturing an interference-containing complex-valued radar data set through simulation, and reserving the interference-free complex-valued radar image data as a supervision tag during training; The system comprises a training data set, a target identification module, an anti-interference and target feature enhancement module, a target identification module and a target identification module, wherein the training data set is used for training a network, and interference-free complex-valued radar image data and target categories are used as supervision labels, so that the network can perform anti-interference and target identification learning at the same time to obtain a trained end-to-end complex-valued neural network; Inputting the complex-valued radar image data to be detected into a trained end-to-end complex-valued neural network to obtain the identification result of the radar target under radio frequency interference. Further, the making of the data set includes: Obtaining interference-free complex-valued radar echo data Acquiring or obtaining radio frequency interference data with different intensities and different categories through simulation; adding radio frequency interference data to raw interference-free radar echo data along the fast time dimension Obtaining radar echo data containing interference; The obtained radar echo data containing interferenceImaging processing is carried out to obtain complex-valued radar image data containing interferenceAt the same time, radar echo data without interference to the originalImaging processing is carried out to obtain interference-free complex-valued radar image data; For complex-valued radar image data containing interference and no interference respectivelyAndThe same segmentation operation is carried out to obtain complex-valued radar image data containing single targ