CN116125421B - Array radar multi-echo signal target detection method based on deep learning - Google Patents
Array radar multi-echo signal target detection method based on deep learning Download PDFInfo
- Publication number
- CN116125421B CN116125421B CN202310073357.1A CN202310073357A CN116125421B CN 116125421 B CN116125421 B CN 116125421B CN 202310073357 A CN202310073357 A CN 202310073357A CN 116125421 B CN116125421 B CN 116125421B
- Authority
- CN
- China
- Prior art keywords
- signal
- array radar
- target
- target detection
- noise
- 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
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
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/417—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section involving the use of neural networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention belongs to the technical field of radar target detection application, and discloses a target detection method of array radar multi-echo signals based on deep learning. And training an array radar signal fusion neural network by adopting the frequency domain feature matrix set to obtain a trained network model. And extracting a frequency domain feature matrix of the array radar detection signal to be detected, carrying out target detection through a network model, judging whether a target exists in the array radar detection signal to be detected, completing target detection, and improving target detection precision.
Description
Technical Field
The invention belongs to the technical field of radar target detection application, and particularly relates to an array radar multi-echo signal target detection method based on a deep learning algorithm under a low signal-to-noise ratio.
Background
When the existing radar detects low/ultra-low-altitude targets, the signal-to-noise ratio of echoes of the low/ultra-low-altitude targets is low due to the influence of electromagnetic interference, multipath interference, beam splitting and other factors, so that the detection and tracking of the radar on the targets in the areas can be influenced to a great extent. Meanwhile, some stealth technologies such as wave-absorbing materials, wave-absorbing structures and shaping at present enable radar reflection sectional areas of stealth targets to be greatly reduced, so that signal-to-noise ratio of echoes of the targets is low, and great difficulty is brought to target detection based on radar signals. In order to improve the combat power of modern radars, the problem of target detection under signal conditions at low signal-to-noise ratios must be solved.
The main method for solving the problem of radar detection capability is to improve the radar transmitting power, reduce the noise coefficient of a receiver, increase the antenna aperture, gain and other hardware measures to improve the signal-to-noise ratio of weak target echoes, but the complexity and cost of hardware design are greatly increased; in another method, a plurality of radars are adopted to perform cooperative detection, and in terms of signal processing, a plurality of radar signals are respectively subjected to fusion processing of received echo signals in the fields of space domain, time domain, frequency domain, polarization domain and the like, so that weak target echo signals annihilated in clutter or noise are enhanced, and are detected. Compared with the first method, the second method has stronger practical value and engineering significance. However, the method has high requirements on the capabilities of multi-signal fusion and target detection classification technologies, and the current industry is in need of designing an array radar multi-echo signal target detection technology with strong robustness and good fusion performance to support the implementation of the second method.
Disclosure of Invention
The invention aims to solve the bottleneck problems of efficient fusion and target detection of multiple signals of an array radar, and provides a target detection method of multiple echo signals of the array radar based on deep learning. Compared with the prior method, the array radar multi-signal fusion neural network designed by the method can replace a plurality of steps such as phase offset correction, signal fusion and target detection of multi-target signals, so that the calculation complexity is effectively reduced, and meanwhile, the target detection precision is further improved.
The invention adopts the technical scheme that:
an array radar multi-echo signal target detection method based on deep learning comprises the following steps:
(1) M target echo signals are synthesized through simulation and serve as a single radar multi-target detection noiseless signal data set; performing N-time phase weighting on each target signal in the single-radar multi-target detection noiseless signal data set to form an array radar multi-target detection noiseless signal data set; white noise with different intensities is respectively added into the array radar multi-target detection noise-free signal data set to obtain S mixed signal data with different signal to noise ratios, and the S mixed signal data are used as the array radar multi-target detection multi-signal to noise ratio data set; wherein M, N and S are set values;
(2) Performing Fourier transform on each signal in the array radar multi-target detection multi-signal-to-noise ratio data set, intercepting the front k low-frequency points as frequency domain features, and combining the frequency features of N radars to form a target frequency feature matrix set of S, M, N, row and k columns; wherein k is a set value;
(3) Generating M x N white noise signals with S different signal intensities, carrying out Fourier transform on each signal, intercepting the front k low-frequency points as noise frequency domain features, and combining the N noise frequency domain features to form a noise frequency feature matrix set of S x M N rows and k columns; mixing the noise frequency characteristic matrix set and the target frequency characteristic matrix set to obtain 2 x S x M frequency characteristic matrix sets of N rows and k columns as training data sets of the array radar multi-signal fusion neural network; the label of the noise frequency characteristic matrix set is set to be 0, and the label of the target frequency characteristic matrix set is set to be 1;
(4) Designing an array radar multi-signal fusion neural network model structure, inputting a training data set into the array radar multi-signal fusion neural network for training, and obtaining a trained array radar multi-signal fusion neural network model;
(5) Carrying out Fourier transform on each signal in the array radar detection signals to be detected, intercepting the front k low-frequency points as frequency domain features to obtain N rows and k columns of frequency feature matrixes to be detected, and inputting the frequency feature matrixes to be detected into a trained array radar multi-signal fusion neural network model to obtain detection result values; if the detection result value is larger than or equal to the set threshold value, the target exists in the array radar detection signal to be detected, and if the detection result value is smaller than the set threshold value, the target does not exist in the array radar detection signal to be detected.
Further, the structure of the array radar multi-signal fusion neural network in the step (4) comprises a dense convolution kernel group, a multi-convolution kernel fusion extraction network and a fully-connected classification network;
the dense convolution kernel group consists of a plurality of convolution kernels with different parameters, wherein each convolution kernel is used for carrying out frequency domain shifting on the frequency characteristics of each radar in the characteristic matrix through convolution operation, adding the characteristics of all the radars according to the frequencies, and obtaining a characteristic set after shifting and adding with different scales through the convolution kernels with different parameters;
the multi-convolution kernel fusion extraction network is used for extracting target features in different scales in the feature set after the shifting and adding of different scales;
the fully-connected classification network is used for acquiring classification results by utilizing the extracted target features under different scales.
Compared with the background technology, the invention has the following advantages:
1. the invention provides a target detection method based on multiple echo signals of an array radar with deep learning, designs an array radar multiple signal fusion neural network, learns the phase shift characteristic among frequency domain characteristics among multiple signals of the array radar in a training mode, corrects and fuses the phase shift characteristic through a characteristic fusion method, and realizes superposition enhancement of weak target signals in multiple low signal-to-noise ratio signals, and realizes target detection while target signal enhancement.
2. Compared with the prior method, the method replaces a plurality of processing procedures such as phase offset correction, signal fusion and target detection of the multi-target signal with a training procedure of a deep neural network structure, effectively reduces the computational complexity, and further improves the target detection precision.
Drawings
FIG. 1 is a block flow frame design of the present invention.
FIG. 2 is a block diagram of an array radar multi-signal fusion neural network of the present invention.
Detailed Description
The following description of the embodiments of the invention is presented in conjunction with the accompanying drawings to provide a better understanding of the invention to those skilled in the art. It is to be expressly noted that in the description below, detailed descriptions of known functions and designs are omitted here as perhaps obscuring the present invention.
FIG. 1 is a schematic flow chart diagram of an implementation of the deep learning-based array radar multi-echo signal target detection method of the present invention.
In this embodiment, as shown in fig. 1, the method for detecting multiple echo signal targets of the array radar based on deep learning includes the following steps:
(1) M target echo signals are synthesized through simulation and serve as a single radar multi-target detection noiseless signal data set; each target signal in the single-radar multi-target detection noiseless signal data set is subjected to N-time phase weighting, and different phase weighting is equivalent to signals of different radars, so that an array radar multi-target detection noiseless signal data set is formed; white noise with different intensities is respectively added into the array radar multi-target detection noiseless signal data set to obtain S mixed signal data with different signal to noise ratios such as 0db, -5db, -10db, -15db and the like, and the S mixed signal data are used as the array radar multi-target detection multi-signal to noise ratio data set;
(2) 10000-point Fourier transform is carried out on each signal in the array radar multi-target detection multi-signal-to-noise ratio dataset, and the first 50 low-frequency points are intercepted to serve as frequency domain features, so that signal feature extraction is completed. Combining the frequency characteristics of N radars to form S.times.M target frequency characteristic matrix sets of N rows and 50 columns;
(3) Generating M x N white noise signals with S different signal intensities, performing 10000-point Fourier transform on each signal, and intercepting the first 50 low-frequency points as noise frequency domain characteristics. And combining the N noise frequency characteristics to form S.M noise frequency characteristic matrix sets of N rows and 50 columns. Mixing the noise frequency characteristic matrix set and the target frequency characteristic matrix set to obtain 2 x S x M N rows and 50 columns of frequency characteristic matrix sets as training data sets of the array radar multi-signal fusion neural network, wherein the label of the noise frequency characteristic matrix set is 0, and the label of the target frequency characteristic matrix set is set to be 1;
(4) Designing an array radar multi-signal fusion neural network model structure, inputting a training data set into the array radar multi-signal fusion neural network for training, and obtaining a trained array radar multi-signal fusion neural network model;
(5) And performing 10000-point Fourier transform on each signal in the radar detection signals of the array to be detected, and intercepting the first 50 low-frequency points as frequency domain features to obtain N rows and 50 columns of frequency feature matrixes to be detected. And inputting the frequency characteristic matrix to be detected into the trained array radar multi-signal fusion neural network model to obtain a detection result value. If the value of the detection result is more than or equal to 0.5, the detection result represents that the target exists in the radar detection signal of the array to be detected. If the value of the detection result is smaller than 0.5, the detection result represents that no target exists in the radar detection signal of the array to be detected.
The structure of the array radar multi-signal fusion neural network in the step (4) comprises a dense convolution kernel group, a multi-convolution kernel fusion extraction network and a fully-connected classification network; as shown in fig. 2.
The dense convolution kernel group consists of a plurality of convolution kernels with different parameters, wherein each convolution kernel is used for carrying out frequency domain shifting on the frequency characteristics of each radar in the characteristic matrix through convolution operation, adding the characteristics of all the radars according to the frequencies, and obtaining a characteristic set after shifting and adding with different scales through the convolution kernels with different parameters;
the multi-convolution kernel fusion extraction network is used for extracting target features in different scales in the feature set after the shifting and adding of different scales;
the fully-connected classification network is used for acquiring classification results by utilizing the extracted target features under different scales.
While the foregoing describes illustrative embodiments of the present invention to facilitate an understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, but is to be construed as protected by the accompanying claims insofar as various changes are within the spirit and scope of the present invention as defined and defined by the appended claims.
Claims (2)
1. The array radar multi-echo signal target detection method based on deep learning is characterized by comprising the following steps of:
(1) M target echo signals are synthesized through simulation and serve as a single radar multi-target detection noiseless signal data set; performing N-time phase weighting on each target signal in the single-radar multi-target detection noiseless signal data set to form an array radar multi-target detection noiseless signal data set; white noise with different intensities is respectively added into the array radar multi-target detection noise-free signal data set to obtain S mixed signal data with different signal to noise ratios, and the S mixed signal data are used as the array radar multi-target detection multi-signal to noise ratio data set; wherein M, N and S are set values;
(2) Performing Fourier transform on each signal in the array radar multi-target detection multi-signal-to-noise ratio dataset, intercepting the front k low-frequency points as frequency characteristics, and combining the frequency characteristics of N radars to form a target frequency characteristic matrix set of S, M, N, row and k columns; wherein k is a set value;
(3) Generating M x N white noise signals with S different signal intensities, carrying out Fourier transform on each signal, intercepting the front k low-frequency points as noise frequency characteristics, and combining the N noise frequency characteristics to form a noise frequency characteristic matrix set of S x M N rows and k columns; mixing the noise frequency characteristic matrix set and the target frequency characteristic matrix set to obtain 2 x S x M frequency characteristic matrix sets of N rows and k columns as training data sets of the array radar multi-signal fusion neural network; the label of the noise frequency characteristic matrix set is set to be 0, and the label of the target frequency characteristic matrix set is set to be 1;
(4) Designing an array radar multi-signal fusion neural network model structure, inputting a training data set into the array radar multi-signal fusion neural network for training, and obtaining a trained array radar multi-signal fusion neural network model;
(5) Carrying out Fourier transform on each signal in the array radar detection signals to be detected, intercepting the front k low-frequency points as frequency characteristics to obtain N rows and k columns of frequency characteristic matrixes to be detected, and inputting the frequency characteristic matrixes to be detected into a trained array radar multi-signal fusion neural network model to obtain detection result values; if the detection result value is larger than or equal to the set threshold value, the target exists in the array radar detection signal to be detected, and if the detection result value is smaller than the set threshold value, the target does not exist in the array radar detection signal to be detected.
2. The method for detecting the array radar multi-echo signal target based on deep learning according to claim 1, wherein the structure of the array radar multi-signal fusion neural network in the step (4) comprises a dense convolution kernel group, a multi-convolution kernel fusion extraction network and a fully-connected classification network;
the dense convolution kernel group consists of a plurality of convolution kernels with different parameters, wherein each convolution kernel is used for carrying out frequency domain shifting on the frequency characteristics of each radar in the frequency characteristic matrix set through convolution operation, adding the characteristics of all the radars according to the frequencies, and obtaining a characteristic set after shifting and adding with different scales through the convolution kernels with different parameters;
the multi-convolution kernel fusion extraction network is used for extracting target features in different scales in the feature set after the shifting and adding of different scales;
the fully-connected classification network is used for acquiring classification results by utilizing the extracted target features under different scales.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310073357.1A CN116125421B (en) | 2023-02-07 | 2023-02-07 | Array radar multi-echo signal target detection method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310073357.1A CN116125421B (en) | 2023-02-07 | 2023-02-07 | Array radar multi-echo signal target detection method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116125421A CN116125421A (en) | 2023-05-16 |
| CN116125421B true CN116125421B (en) | 2023-10-20 |
Family
ID=86304368
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310073357.1A Active CN116125421B (en) | 2023-02-07 | 2023-02-07 | Array radar multi-echo signal target detection method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116125421B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117572376B (en) * | 2024-01-16 | 2024-04-19 | 烟台大学 | Low signal-to-noise ratio weak and small target radar echo signal recognition device and training recognition method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110471026A (en) * | 2019-07-22 | 2019-11-19 | 西安电子科技大学 | A kind of low elevation angle DOA estimation method of metre wave radar target of phase enhancing |
| CN111693946A (en) * | 2019-03-14 | 2020-09-22 | 英飞凌科技股份有限公司 | FMCW radar with interference signal suppression by means of an artificial neural network |
| CN112612005A (en) * | 2020-11-27 | 2021-04-06 | 中山大学 | Radar main lobe interference resisting method based on deep learning |
| CN113486961A (en) * | 2021-07-12 | 2021-10-08 | 安徽耀峰雷达科技有限公司 | Radar RD image target detection method and system based on deep learning under low signal-to-noise ratio and computer equipment |
| CN115236584A (en) * | 2022-06-10 | 2022-10-25 | 中国人民解放军空军工程大学 | Meter-wave radar low elevation angle estimation method based on deep learning |
| CN115372925A (en) * | 2022-08-10 | 2022-11-22 | 中山大学 | Array robust adaptive beam forming method based on deep learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US12013484B2 (en) * | 2020-05-20 | 2024-06-18 | Infineon Technologies Ag | Radar receiving system and method for compensating a phase error between radar receiving circuits |
-
2023
- 2023-02-07 CN CN202310073357.1A patent/CN116125421B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111693946A (en) * | 2019-03-14 | 2020-09-22 | 英飞凌科技股份有限公司 | FMCW radar with interference signal suppression by means of an artificial neural network |
| CN110471026A (en) * | 2019-07-22 | 2019-11-19 | 西安电子科技大学 | A kind of low elevation angle DOA estimation method of metre wave radar target of phase enhancing |
| CN112612005A (en) * | 2020-11-27 | 2021-04-06 | 中山大学 | Radar main lobe interference resisting method based on deep learning |
| CN113486961A (en) * | 2021-07-12 | 2021-10-08 | 安徽耀峰雷达科技有限公司 | Radar RD image target detection method and system based on deep learning under low signal-to-noise ratio and computer equipment |
| CN115236584A (en) * | 2022-06-10 | 2022-10-25 | 中国人民解放军空军工程大学 | Meter-wave radar low elevation angle estimation method based on deep learning |
| CN115372925A (en) * | 2022-08-10 | 2022-11-22 | 中山大学 | Array robust adaptive beam forming method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116125421A (en) | 2023-05-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Melvin et al. | Knowledge-aided signal processing: a new paradigm for radar and other advanced sensors | |
| CN109407055B (en) | Beam forming method based on multipath utilization | |
| CN106842140B (en) | A kind of main lobe interference suppression method based on difference beam dimensionality reduction | |
| CN109444869B (en) | Radar extension target parameter adjustable detector for signal mismatch | |
| CN112612005B (en) | Radar main lobe interference resistance method based on deep learning | |
| CN116125421B (en) | Array radar multi-echo signal target detection method based on deep learning | |
| CN109444820A (en) | Object detection method after Multichannel radar elder generation AF panel when clutter and interference co-existence | |
| CN113238211B (en) | Parameterized adaptive array signal detection method and system under interference condition | |
| CN101644760A (en) | Rapid and robust method for detecting information source number suitable for high-resolution array | |
| CN111257879A (en) | Method for solving millimeter wave MIMO radar target splitting based on two norms | |
| CN115372925A (en) | Array robust adaptive beam forming method based on deep learning | |
| CN107748364A (en) | Low wind field speed estimation method based on contraction multistage wiener filter | |
| CN110361697B (en) | Robust beam forming method based on covariance matrix hybrid reconstruction | |
| CN110146854B (en) | Robust anti-interference method for FDA-MIMO radar | |
| CN115575921B (en) | Pitching-direction-based multichannel multi-interference-base suppression interference suppression method | |
| CN111044996A (en) | LFMCW radar target detection method based on dimension reduction approximate message transfer | |
| CN114152918B (en) | Anti-intermittent main lobe interference method based on compressed sensing | |
| CN111931570B (en) | Through-wall imaging radar human body target detection method based on full convolution network | |
| CN115524672A (en) | Target robustness intelligent detection method under structured interference and clutter | |
| CN110346766B (en) | Null broadening method based on sparse constraint control side lobe | |
| CN113406578A (en) | Target detection method and device for distributed unmanned airborne radar and storage medium | |
| CN113030928A (en) | Polarization radar extended target self-adaptive detection method and system in non-uniform environment | |
| Feng et al. | Constained adaptive monopulse algorithm based on sub-array | |
| CN114527444B (en) | Airborne MIMO radar self-adaptive clutter suppression method based on space-time sampling matrix | |
| CN115685081B (en) | GLRT-based method for detecting distance expansion target in interference plus noise background |
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 |