CN111983555B - Interferometer angle resolving method based on neural network - Google Patents
Interferometer angle resolving method based on neural network Download PDFInfo
- Publication number
- CN111983555B CN111983555B CN202010791037.6A CN202010791037A CN111983555B CN 111983555 B CN111983555 B CN 111983555B CN 202010791037 A CN202010791037 A CN 202010791037A CN 111983555 B CN111983555 B CN 111983555B
- Authority
- CN
- China
- Prior art keywords
- interferometer
- network
- angle
- phase
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S3/00—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received
- G01S3/02—Direction-finders for determining the direction from which infrasonic, sonic, ultrasonic, or electromagnetic waves, or particle emission, not having a directional significance, are being received using radio waves
- G01S3/14—Systems for determining direction or deviation from predetermined direction
- G01S3/38—Systems for determining direction or deviation from predetermined direction using adjustment of real or effective orientation of directivity characteristic of an antenna or an antenna system to give a desired condition of signal derived from that antenna or antenna system, e.g. to give a maximum or minimum signal
- G01S3/40—Systems for determining direction or deviation from predetermined direction using adjustment of real or effective orientation of directivity characteristic of an antenna or an antenna system to give a desired condition of signal derived from that antenna or antenna system, e.g. to give a maximum or minimum signal adjusting orientation of a single directivity characteristic to produce maximum or minimum signal, e.g. rotatable loop antenna or equivalent goniometer system
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Variable-Direction Aerials And Aerial Arrays (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to an interferometer angle resolving method based on a neural network algorithm, which is characterized in that an interferometer angle equation curve is fitted through a multi-layer neural network, so that the arrival angle of an incoming wave signal can be calculated according to input frequency and phase information. The invention can train aiming at a single interferometer independently, thereby omitting the correction process of the interferometer and improving the angle measurement precision. The invention has the characteristic of insensitivity to the interferometer array mode, and can adapt to a plurality of different interferometer array modes through the same training mode. Meanwhile, the invention is insensitive to the angle range of the incoming wave signal, and can fit a nonlinear curve under the condition of large angle, so that the angle measurement range of the interferometer can be expanded.
Description
Technical Field
The invention belongs to the technical field of electronic countermeasure data processing, in particular to an angle calculation method for an interferometer electronic reconnaissance system, and further relates to an antenna array mode and a calibration method for the interferometer electronic reconnaissance system, so as to simplify the calibration flow of the interferometer electronic reconnaissance system, expand the array mode and improve the direction finding precision.
Background
Electronic reconnaissance systems have become an important information acquisition means for modern war by using fast, high-precision and high-recognition passive detection technology to perform battlefield monitoring and situation awareness. The interferometer equipment has the capability of positioning and direction-finding the signal source, has the advantages of high system precision, high sensitivity, strong anti-interference performance, short data processing time, wide working frequency range and the like, is widely used, and is used as an important component part of an electronic reconnaissance system, and the main function is as follows: 1, sorting and identification of radiation sources provide reliable basis (the spatial position of the radiation sources cannot be changed quickly); 2, measuring the azimuth angle of the radiation source with high precision and further implementing passive positioning; 3 provide guidance for electronic interference and destruction attacks. The interferometer direction finding system uses the phase difference of the signals between the antennas of the electromagnetic wave arrival direction finding antenna array to further estimate the incoming wave path difference, and determines the incoming wave direction according to the antenna spacing and the wave path difference.
According to interferometer direction-finding principle, its direction-finding accuracy and phase measurement accuracy and antenna array mode are related.
The phase measurement accuracy of the interferometer system is mainly influenced by antenna phase characteristics, receiver phase consistency, phase jitter caused by noise and the like, and in the traditional interferometer angle measurement method, under the condition of determining a base line size, the phase jitter and the angle measurement accuracy are approximately in a linear relation.
The conventional interferometer system array mode directly determines an angle calculation algorithm, namely an angle calculation formula depends on the antenna array mode and the array parameters. Under different application scenes, the angle calculation formula and parameters need to be deduced again to change the array mode and parameters.
In actual direction finding, the problem of multi-value ambiguity also occurs, and a true phase difference needs to be found by using an ambiguity resolution algorithm. The deblurring algorithm adopted by the traditional interferometer needs to take the approximate linear relation of the angle and the phase difference as the premise, the approximate linear relation only exists under small angle measurement, and under the condition of large angle measurement, the correlation of the phase difference and the angle is not linear any more, so that the deblurring accuracy of the traditional interferometer in a nonlinear angle measurement area is seriously reduced.
In summary, the conventional interferometer angle calculation method has three drawbacks:
[1] the lack of effective suppression of phase jitter or a phase jitter adaptation algorithm based on statistics and fitting, so that the angle measurement precision can be directly influenced by the phase jitter intensity of the system;
[2] the antenna array form is single, namely the adaptability to the application scene is single, the theoretical calculation and verification work required when the antenna array needs to be changed is more, and the adaptation to various array modes cannot be realized through a unified algorithm;
[3] the adaptive deblurring angle range is smaller, and the accuracy is seriously reduced when the deblurring is carried out at a large angle.
Disclosure of Invention
Technical problem to be solved
In order to solve the problems that the existing interferometer has more correction parameters under the condition of large bandwidth and the angle measurement precision is seriously deteriorated due to phase jitter, the invention provides an interferometer angle resolving method based on a neural network.
Technical proposal
An interferometer angle resolving method based on a neural network is characterized by comprising the following steps:
step 1: acquiring phase information under a determined angle by using an interferometer; preprocessing the obtained information:
s101, in a microwave darkroom, an interferometer is arranged on an experimental turntable, the angle between the interferometer and a signal source is adjusted by the turntable, the turntable is adjusted at certain angle intervals, the phase and the frequency of a signal are measured by the interferometer, four different phase values are obtained by 4 measuring antennas of the interferometer, and the same signal corresponds to a frequency value f;
s102, deleting some phase data which are obvious interference signals according to priori information;
s103, 4 antennas of the interferometer are distributed transversely, phase differences of the antennas at adjacent positions are calculated, and 3 groups of phase difference data PH= [ PH ] are obtained 1 ,PH 2 ,PH 3 ]And (3) carrying out normalization processing on the phase difference data and the frequency:
step 2: selecting a fully connected network as a training model, wherein the phase difference and the frequency value data= [ PH, f ] correspondingly input by the input nodes, so that the number of the nodes of the input layer is set to be 4, and the corresponding angles of the neurons are output; the number of network layers can be set to 4-10 layers, and the number of neurons in each layer can be set to 20-200; during training, a mean square error function of a network predicted value and a true value is selected as a loss function of the network:
wherein Ang i The angle of the turntable is taken as a known signal angle, namely a true value; ang i ' is a network predicted value;
step 3: randomly initializing weights in a network, inputting training data into the network for network training, wherein the input data can be trained in batches according to the data quantity, and the network training is performed by a gradient-based back propagation algorithm until a loss function converges;
step 4: the observed phase information and signal frequency of the interferometer are input into a trained network after data preprocessing, and a model is utilized to carry out forward operation so as to obtain a predicted value of the signal angle.
The certain angle in step 1 is 1 °.
The fully connected network described in step 2 may be replaced by a recursive network, a residual network.
Advantageous effects
According to the interferometer angle resolving method based on the neural network, through the fuzzy resolving and angle resolving process of the multi-layer neural network fitting interferometer, an angle computing parameter matrix aiming at each interferometer product is generated by a neural network training method with reasonable design, and correction parameters and angle computing equations of the existing interferometer phase are replaced; meanwhile, when the angle calculation is carried out by adopting the parameter matrix generated after the neural network training, the calculated amount is the multiplication and addition operation of the determined times, which is equivalent to the calculated amount of the prior method, and the real-time performance of the interferometer in the use process is not affected. Compared with the prior art, the method has the following beneficial effects:
1. the training is performed for each interferometer product in a customized manner through the neural network, in the training process, the initial phase is trained in a parameterized manner, the work of initial phase correction is omitted, meanwhile, the customized training process can accurately fit the phase-angle of arrival curve of each interferometer product, and as the neural network has more layers, a higher-order function can be fitted, so that the phase jitter and the angle measurement error are not in a simple linear relationship any more, but are closer to the phase characteristic of a single interferometer through the fitted higher-order function curve, and the angle measurement precision can be improved to a certain extent.
2. Compared with the prior interferometer angle measurement algorithm, the invention adopts the input phase value and the arrival angle information in the training process, and the algorithm itself trains and fits the phase-arrival angle function, so that the relation equation of the phase and the arrival angle under the condition of the array is not required to be calculated when aiming at a plurality of typical array modes, and therefore, in actual use, the algorithm is insensitive to the array modes and can fit the angle measurement algorithm of a plurality of typical array modes.
3. Because the angle measurement curve fitted by the multi-layer neural network is a high-order nonlinear function, the nonlinear region fitting can be performed on the large-angle measurement in the existing angle measurement equation, so that the algorithm provided by the invention can realize a larger angle measurement range than the existing angle calculation method, and in the range, the angle measurement precision cannot be greatly deteriorated.
Drawings
FIG. 1 is a block diagram of the algorithm training of the present invention
FIG. 2 is a schematic diagram of a neural network according to the present invention
Detailed Description
The invention will now be further described with reference to examples, figures:
the embodiment of the invention mainly comprises three stages: neural network model construction, network model training and network model practical application. The specific implementation steps are as follows:
s1, collecting training data:
the collection of training data mainly includes: acquiring phase information under a determined angle by using an interferometer; the obtained information is preprocessed to be suitable for the later neural network training.
S101, installing an interferometer on an experimental turntable, adjusting the angle between the interferometer and a signal source by using the turntable, adjusting the turntable at 1 degree intervals, measuring the phase and the frequency of a signal by using the interferometer, and obtaining four different phase values by using four measuring antennas, wherein the same signal corresponds to one frequency value f. The angle of the turntable is taken as the known signal angle (Ang). To ensure the training accuracy, the training process is recommended to be carried out in a microwave darkroom.
S102, deleting some phase data which are obvious interference signals according to the prior information.
S103, obtaining phase difference data from the phase data. 4 antennas of the interferometer are transversely distributed, phase differences of the antennas at adjacent positions are calculated, and 3 groups of phase difference data PH= [ PH ] can be obtained 1 ,PH 2 ,PH 3 ]The phase difference data is normalized, and the frequency is normalized, and the specific operation is as follows:
s2, building neural network model
The stage mainly builds a proper neural network model according to the actual interferometer angle settlement problem.
S201, selecting a basic network architecture. A basic fully connected network may be selected, as shown in fig. 2, as well as other network architectures, such as a recursive network, a residual network, etc.
S202, designing a network specific structure. The input node corresponds to the input phase difference and frequency value data= [ PH, f ], so the input layer node number is set to 4, the output neuron corresponds to the angle, and according to the representation method of the angle, the input layer node number can be set to 1 neuron or two neurons. The number of network layers can be set to 4-10 layers, and the number of neurons in each layer can be set to 20-200.
S203, setting a network loss function, wherein the invention mainly aims at the relation between the phase difference and the angle, so that a mean square error function of a network predicted value and a true value is selected as the loss function of the network.
S3, neural network model training
S301, selecting a model training algorithm, wherein the invention selects the commonly used gradient-based back propagation algorithm to train the network. After the loss function has converged steadily, the training process of the network ends.
S302, randomly initializing the weight in the network.
S303, inputting training Data into a network for network training, wherein the input Data can be trained in batches according to the size of Data volume, and the training speed is improved under the condition that the effect is not reduced, and if GPU equipment is provided, parallel acceleration can be performed.
S4, practical application of neural network model
In the previous stage, a trained neural network model can be obtained through data acquisition, network construction and training, and parameters in the model are well fit with complex relations between measured data (phase and frequency) of the interferometer and angles of signal sources. In practical application, only the test data (including phase and signal frequency) of the interferometer is required to be preprocessed, then the phase difference and the frequency value are input into a trained model, and the model is utilized to carry out forward operation to obtain the predicted value of the signal angle.
Claims (3)
1. An interferometer angle resolving method based on a neural network is characterized by comprising the following steps:
step 1: acquiring phase information under a determined angle by using an interferometer; preprocessing the obtained information:
s101, in a microwave darkroom, an interferometer is arranged on an experimental turntable, the angle between the interferometer and a signal source is adjusted by the turntable, the turntable is adjusted at certain angle intervals, the phase and the frequency of a signal are measured by the interferometer, four different phase values are obtained by 4 measuring antennas of the interferometer, and the same signal corresponds to a frequency value f;
s102, deleting some phase data which are obvious interference signals according to priori information;
s103, 4 antennas of the interferometer are distributed transversely, phase differences of the antennas at adjacent positions are calculated, and 3 groups of phase difference data PH= [ PH ] are obtained 1 ,PH 2 ,PH 3 ]And (3) carrying out normalization processing on the phase difference data and the frequency:
step 2: selecting a fully connected network as a training model, wherein the phase difference and the frequency value data= [ PH, f ] correspondingly input by the input nodes, so that the number of the nodes of the input layer is set to be 4, and the corresponding angles of the neurons are output; the number of the network layers is set to be 4-10, and the number of the neurons in each layer is set to be 20-200; during training, a mean square error function of a network predicted value and a true value is selected as a loss function of the network:
wherein Ang i The angle of the turntable is taken as a known signal angle, namely a true value; ang i ' is a network predicted value;
step 3: randomly initializing weights in a network, inputting training data into the network for network training, carrying out batch training on the input data according to the data quantity, and carrying out network training on the basis of a gradient back propagation algorithm until a loss function converges;
step 4: the observed phase information and signal frequency of the interferometer are input into a trained network after data preprocessing, and a model is utilized to carry out forward operation so as to obtain a predicted value of the signal angle.
2. The method of claim 1, wherein the certain angle in step 1 is 1 °.
3. The method for resolving angles of interferometers based on neural network according to claim 1, wherein the fully connected network in step 2 is replaced by a recursive network or a residual network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010791037.6A CN111983555B (en) | 2020-08-07 | 2020-08-07 | Interferometer angle resolving method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010791037.6A CN111983555B (en) | 2020-08-07 | 2020-08-07 | Interferometer angle resolving method based on neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111983555A CN111983555A (en) | 2020-11-24 |
| CN111983555B true CN111983555B (en) | 2023-07-11 |
Family
ID=73445893
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010791037.6A Active CN111983555B (en) | 2020-08-07 | 2020-08-07 | Interferometer angle resolving method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111983555B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100077861A (en) * | 2008-12-29 | 2010-07-08 | 성균관대학교산학협력단 | Compensating apparatus and method for heterodyne laser interferometer |
| US9229102B1 (en) * | 2009-12-18 | 2016-01-05 | L-3 Communications Security And Detection Systems, Inc. | Detection of movable objects |
| CN108494602A (en) * | 2018-04-08 | 2018-09-04 | 上海鸿洛通信电子有限公司 | Method of adjustment, device and the intelligent terminal of OTA parameters |
| CN108932480A (en) * | 2018-06-08 | 2018-12-04 | 电子科技大学 | The study of distributing optical fiber sensing signal characteristic and classification method based on 1D-CNN |
| CN109683158A (en) * | 2018-12-21 | 2019-04-26 | 西安电子工程研究所 | The static antiaircraft unit detecting method in radar ground based on the fusion of interferometer information |
| CN110260889A (en) * | 2019-07-11 | 2019-09-20 | 中国计量科学研究院 | A kind of hand-hold ranger automatic calibrator and its calibration method |
| CN110470263A (en) * | 2019-08-02 | 2019-11-19 | 天津大学 | A kind of revolving body measurement system error compensation method based on gradient boosted tree |
| CN111044971A (en) * | 2019-12-23 | 2020-04-21 | 南京长峰航天电子科技有限公司 | Two-dimensional interferometer direction finding system |
| CN111487478A (en) * | 2020-03-27 | 2020-08-04 | 杭州电子科技大学 | Angle-dependent complex array error calibration method based on deep neural network |
-
2020
- 2020-08-07 CN CN202010791037.6A patent/CN111983555B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100077861A (en) * | 2008-12-29 | 2010-07-08 | 성균관대학교산학협력단 | Compensating apparatus and method for heterodyne laser interferometer |
| US9229102B1 (en) * | 2009-12-18 | 2016-01-05 | L-3 Communications Security And Detection Systems, Inc. | Detection of movable objects |
| CN108494602A (en) * | 2018-04-08 | 2018-09-04 | 上海鸿洛通信电子有限公司 | Method of adjustment, device and the intelligent terminal of OTA parameters |
| CN108932480A (en) * | 2018-06-08 | 2018-12-04 | 电子科技大学 | The study of distributing optical fiber sensing signal characteristic and classification method based on 1D-CNN |
| CN109683158A (en) * | 2018-12-21 | 2019-04-26 | 西安电子工程研究所 | The static antiaircraft unit detecting method in radar ground based on the fusion of interferometer information |
| CN110260889A (en) * | 2019-07-11 | 2019-09-20 | 中国计量科学研究院 | A kind of hand-hold ranger automatic calibrator and its calibration method |
| CN110470263A (en) * | 2019-08-02 | 2019-11-19 | 天津大学 | A kind of revolving body measurement system error compensation method based on gradient boosted tree |
| CN111044971A (en) * | 2019-12-23 | 2020-04-21 | 南京长峰航天电子科技有限公司 | Two-dimensional interferometer direction finding system |
| CN111487478A (en) * | 2020-03-27 | 2020-08-04 | 杭州电子科技大学 | Angle-dependent complex array error calibration method based on deep neural network |
Non-Patent Citations (4)
| Title |
|---|
| 一种相位干涉仪的数字信道化测向方法;余建宇 等;火控雷达技术;第40卷(第1期);全文 * |
| 基于RBF神经网络的相关干涉仪测向方法;赵雷鸣 等;无线电工程;第41卷(第1期);全文 * |
| 基于神经网络的干涉仪测向方法;崔功杰 等;无线电工程;第43卷(第2期);全文 * |
| 数控XY工作台动态定位误差分析与建模;庞勇军;中国优秀硕士学位论文全文数据库;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111983555A (en) | 2020-11-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105785339A (en) | Airborne radar clutter covariance matrix estimation method in inhomogeneous clutter environment | |
| CN108957403B (en) | Gaussian fitting envelope time delay estimation method and system based on generalized cross correlation | |
| CN109581281B (en) | Moving target positioning method based on arrival time difference and arrival frequency difference | |
| CN111999716B (en) | Clutter prior information-based target adaptive fusion detection method | |
| CN104023394A (en) | WSN positioning method based on self-adaptation inertia weight | |
| CN111044975A (en) | Method and system for positioning earth vibration signal | |
| CN110007299B (en) | Weak target detection tracking method based on mixed coordinate pseudo-spectrum technology | |
| CN109212466B (en) | Quantum dragonfly evolution mechanism-based broadband direction finding method | |
| CN107356908B (en) | Frequency agile signal coherent accumulation method | |
| CN111624549B (en) | Passive filtering tracking method under non-common-view condition | |
| CN105676169B (en) | A kind of orientation method and device of pulse acoustic target | |
| CN112014835B (en) | Target tracking method and device of distributed sparse array radar under grating lobe ambiguity | |
| CN114114240B (en) | Three-dimensional target tracking method and device of ultra-sparse array under influence of grating lobes | |
| CN111999714A (en) | Self-adaptive fusion detection method based on multi-scattering point estimation and clutter knowledge assistance | |
| Zhang et al. | A three-dimensional node security localization method for WSN based on improved RSSI-LSSVR algorithm | |
| NL2026449B1 (en) | Abnormal grain condition detection method based on radio tomographic imaging | |
| CN115113155A (en) | Airborne distributed aperture coherent synthetic radar testing and evaluating method | |
| CN111983555B (en) | Interferometer angle resolving method based on neural network | |
| CN111505566B (en) | Ultrahigh frequency radio frequency signal DOA estimation method | |
| CN117517892A (en) | Transformer partial discharge ultrahigh frequency positioning method and system considering non-line-of-sight propagation | |
| CN110967675A (en) | Passive detection method for satellite external radiation source target based on neural network | |
| CN103218509A (en) | Composite anti-interference data association method based on hardware | |
| Ou et al. | An improved node localization based on adaptive iterated unscented Kalman filter for WSN | |
| CN117908034B (en) | Robust modular basis DOA estimation method based on adaptive beam tracking underwater target | |
| Wu et al. | Study on the performance of the variable step-size LMS algorithms |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |