CN114609602B - Feature extraction-based target detection method under sea clutter background - Google Patents
Feature extraction-based target detection method under sea clutter background Download PDFInfo
- Publication number
- CN114609602B CN114609602B CN202210229272.3A CN202210229272A CN114609602B CN 114609602 B CN114609602 B CN 114609602B CN 202210229272 A CN202210229272 A CN 202210229272A CN 114609602 B CN114609602 B CN 114609602B
- Authority
- CN
- China
- Prior art keywords
- beta
- target
- sea clutter
- feature extraction
- vector
- 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
- 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
- G01S7/414—Discriminating targets with respect to background clutter
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- 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)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a method for detecting targets under a sea clutter background based on feature extraction, which is applied to the technical field of target detection in a radar system and aims at solving the problems that the existing method for detecting the targets on the sea surface under the sea clutter background is mainly realized by statistical model fitting or nonlinear signal processing, has the problems of model mismatch, complex realization mode, large computation amount and the like, and is not beneficial to the detection of the targets on the sea surface; the invention utilizes different energy characteristics of the target and the clutter in the radar echo in the time-frequency domain to carry out feature extraction, and then uses the improved support vector machine classifier to classify the target and the clutter, thereby being capable of obviously improving the constant false alarm detection performance.
Description
Technical Field
The invention belongs to the technical field of target detection in a radar system, and particularly relates to a target detection technology under a sea clutter background.
Background
In recent years, with the lightening and stealth of sea surface targets, more and more sea surface targets appear in the field of radar sea detection. Typical sea targets are small vessels, frogs, debris, and submarine periscopes. Such targets generally have the characteristics of low moving speed, small geometric dimension, strong stealth performance and the like, and bring great challenges to the warning and detection of marine radars. The traditional method for improving the detection capability of the marine radar on the sea surface target has the measures of changing the parameters of the radar system by adopting high Doppler, high distance resolution and the like. However, such methods typically require higher development costs to implement. Relatively speaking, the time-frequency characteristics of the sea clutter can be analyzed in a key manner from the angle of the time-frequency domain, and then the sea clutter and the targets are classified by utilizing different characteristics of the sea clutter and the target signals expressed on the time-frequency domain, so that the target detection is realized.
At present, target detection research under a sea clutter background can be mainly divided into the following three categories: the first type is statistical modeling based on sea clutter, the second type is nonlinear characteristics based on sea clutter, and the third type is characteristic space constructed based on different characteristics of sea clutter in different transform domains. Based on statistical theory, the proper model is generally required to be selected to model the distribution of the sea clutter, so that the statistical model can better fit the internal structure of the sea clutter. Common sea clutter statistical models include gaussian distribution, rayleigh distribution, log-normal distribution, weibull distribution, composite K distribution, and the like. With the development of radar technology, radar echo data acquired under the conditions of high resolution and low grazing angle show non-uniform, non-linear and non-stable random characteristics, and under the condition, the research based on the statistical model cannot well fit the distribution of the sea clutter and reflect the physical mechanism of the sea clutter. The research on the nonlinear characteristics of the sea clutter is mainly developed from the chaos and fractal characteristics of the sea clutter, the sea clutter has the multi-fractal characteristics within a certain time scale, the target detection can be realized by calculating the Hurst index of the sea clutter, however, the fractal characteristics exist in a sea clutter signal only in a scale-free area of a certain time scale, and when the observation time is short, the performance of the detector will seriously slide down. The sea surface target detection method based on the transform domain features mainly describes the difference between a sea clutter signal and a target signal from a plurality of angles such as a time domain, a frequency domain, a time-frequency domain, a polarization domain and the like, and then a high-dimensional space is constructed on the basis, and the target detection problem under the sea clutter background is converted into a two-classification problem, so that the target signal and the clutter signal are identified.
In general, the existing sea surface target detection method under the sea clutter background is mainly realized through statistical model fitting or nonlinear signal processing, and the problems of model mismatch, complex realization mode, large computation amount and the like exist, so that the detection of the sea surface target is not facilitated.
Disclosure of Invention
In order to solve the technical problems, the invention provides a method for detecting targets under a sea clutter background based on feature extraction, which is characterized in that different energy characteristics of the targets and the clutter in radar echoes in a time-frequency domain are utilized to extract features, and then an improved support vector machine classifier is used for classifying the targets and the clutter, so that the constant false alarm detection performance can be obviously improved.
The technical scheme adopted by the invention is as follows: a method for detecting targets under a sea clutter background based on feature extraction comprises the following steps:
s1, sampling in a sliding mode from an echo data starting point in a pulse direction by using a sliding window method, and storing a vector after each sliding window as an original data sample vector;
s2, carrying out time-frequency transformation on the original data sample vector,
s3, extracting a low-frequency intrinsic mode function according to the energy distribution characteristic of the time-frequency transformation result;
s4, constructing a new data sample vector according to the extracted low-frequency eigenmode function;
s5, dividing the new data sample vector into a training set and a testing set;
s6, training the support vector machine by using a training set to obtain a weight vector and an offset of an optimal hyperplane;
and S7, testing the weight vector of the optimal hyperplane and the support vector machine under the bias by using the test set to obtain a target detection result.
The invention has the beneficial effects that: the width of the sliding window is larger than the step length of the sliding window every time, so that the obtained adjacent data samples are partially overlapped, and the expansion of the data samples can be realized; the method performs feature extraction by using different energy distribution characteristics of target signals and clutter signals in radar echoes after time-frequency transformation, performs secondary classification on the targets and the clutter by using an improved support vector machine classifier, and can remarkably improve the detection performance of the constant false alarm.
Drawings
FIG. 1 is a general flow chart of an implementation of the present invention;
fig. 2 is radar echo data received by an IPIX radar according to an embodiment of the present invention;
fig. 3 is a low-frequency IMF energy ratio of the target unit and the sea clutter unit after binary empirical mode decomposition according to the embodiment of the present invention;
FIG. 4 is a graph comparing the detection results of the method of the present invention and the conventional CA-CFAR method.
Detailed Description
The invention adopts scientific calculation software Matlab R2020a to process the IPIX data set so as to verify the correctness and the effectiveness of the invention. The distance unit where the main target is located in the sea clutter data is 7, the radar transmitting carrier frequency is 9.3GHz, the sampling frequency is 10MHz, the pulse width is 200ns, and the pulse repetition frequency is 1000Hz. The following provides a specific implementation of the present invention with reference to fig. 1:
step 1: baseband echo preprocessing: loading IPIX sea clutter data by using Matlab R2020a, removing I, Q channel mean value and standard deviation and eliminating channel imbalance to obtain preprocessed two-dimensional baseband echo data which is recorded as z (t, t) n ) Where t denotes the fast time, t n Indicating a slow time. Let t =2r/c and n = t n T, then the range-pulse two-dimensional baseband echo data is z (r, N), where r is the range variable corresponding to T, N (N e [0,1., N)]) Indicating the number of pulses, N being the total number of pulses. Let r = α c/2f s Then the range bin-pulse two-dimensional baseband echo data is z (α, n), where α is the range bin from r, f s Representing the sampling frequency and c the speed of light. As shown in fig. 2, range unit-pulse two-dimensional baseband echo data.
Step 2: raw data sample generation: and extracting column vectors from the baseband echo along a distance dimension to obtain pulse dimension echo data, and then performing sliding sampling from the starting point of the pulse dimension echo data according to a pulse direction by using a sliding window method, wherein the step length of a sliding window is delta N =100, and the width of the sliding window is N =4096. The vector intercepted after each sliding window is stored as a data sample, and because the width of each sliding window is larger than the step length of the sliding window, the obtained adjacent data samples are partially overlapped, thereby realizing the expansion of the data samples. Each sample can be expressed as
z i (α,n)＝z(△n·(q-1)+1:△n·(q-1)+N),q＝1,2....N total
And step 3: binary empirical mode decomposition and low-frequency IMF energy calculation: and carrying out binary empirical mode decomposition on the data sample vector stored after each sliding window to obtain a plurality of intrinsic mode functions and a residual value. For the vector corresponding to the alpha-th distance unit, the result of binary empirical mode decomposition is
Wherein, c i (n) is the ith IMF, r (n) is the residual value, c i Both (n) and r (n) are complex variables.
After each decomposition, the signal energy of each IMF is calculated as
And recorded. Then, the energy ratio of the ith IMF to all IMFs is calculated and recorded as T i The expression is
Based on the properties of the IMF, the T corresponding to the signal of the IMF at the earlier stage is known i Greater value, and later IMF signal corresponding to T i The values are sequentially decremented.
Finding the difference between the energy ratios of adjacent IMFs, i.e. P i ＝T i+1 -T i (I = 1.., I-1). Finding P i Maximum and its index i = k. Since the energy fraction of the low-frequency IMF increases when a target is present in the echo, an IMF with a position index i ≧ k can be considered as the low-frequency IMF. And calculating the energy of the low-frequency IMF and adding to obtain a new data sample vector E (n), thereby obtaining a training set and a data set of the SVM classifier. The vector of data samples E (n) is calculated as
All new data sample vectors E (n) are then randomly drawn in proportion to 8:2 to form a training set and a testing set. After the binary empirical mode decomposition, the low-frequency IMF energy proportion of the partial sample is shown in fig. 3, and it can be seen that the low-frequency IMF energy proportion of the target signal after the time-frequency transform is higher than that of the clutter signal.
And 4, step 4: training and detecting a classifier:
in the conventional support vector machine, samples to be classified share a penalty factor β, which means that the sea clutter signal and the target signal have the same tolerance to outliers in the signal samples. However, in the process of detecting a target on the sea surface, an outlier sample is usually caused by an abnormal factor such as a sea spike, and the influence of the sea spike on the target signal and the sea clutter signal is usually different. Therefore, to solve this problem, the present invention introduces different penalty factors β to the sea clutter signal and the target signal in the sample 0 And beta 1 Instead of using the penalty factor β in the conventional support vector machine to control the error weight of each component in the optimization problem, the mathematical description of the improved support vector machine is as follows:
s.t.ξ i ≥0i＝1,2,...N
y i [k(ω,F i )-b]≥1-ξ i i＝1,2,...N
wherein ξ i For the relaxation variables, ω and b represent the weight vector and the offset, β, respectively, of the optimal hyperplane 0 And beta 1 Respectively representing penalty factors for the sea clutter samples and for the target samples,
is a Gaussian kernel function, F i Represents one sample in the training set and,n represents the number of samples. />
When beta is 0 Or beta 1 When the sampling time is changed, the hyperplane determined by omega and b can incline towards the target sample or the clutter sample, and finally the optimal hyperplane is converged, so that effective classification of the clutter and the target sample is realized. Continuously iterating beta in solving omega and b 0 And beta 1 . The specific iterative process is as follows: firstly, presetting the false alarm probability P to be achieved fa A threshold psi, a penalty factor beta 0 And beta 1 、β 0 Is a value range boundary beta l And beta r . The initial value of the iteration is typically an empirical value, β 0 And beta 1 Usually 1 can be taken, the threshold psi can usually be 0.001, beta l Usually, it may be taken as 0, beta r Usually 2 can be taken. Then, the object and the clutter samples are classified by utilizing the hyperplane determined by omega and b, and the current false alarm probability P is calculated F If P is F And the false alarm probability P to be achieved fa Satisfy | P F -P fa |>Psi, beta can be adjusted according to two conditions 0 And beta 1 And continuing iteration, specifically: if P F <P fa Then set beta r ＝β 0 And beta is 0 ＝(β r +β l ) 2, continuing to iterate; if P is F >P fa Then, set up β l ＝β 0 And beta is 0 ＝(β r +β l ) And/2, continuing the iteration. After each iteration, a new set of ω and b is obtained until | P F -P fa Stopping iteration when | ≦ ψ, wherein ω and b at this time are parameters of the optimal hyperplane.
Reuse test set sample F j And testing the improved support vector machine model under the optimal parameters omega and b to obtain a detection result under the current false alarm probability. The classification criterion is:
wherein, y j =1 denotes that the decision result of the support vector machine is the target sample, y j = -1 indicates that the support vector machine determination result is a sea clutterA wave sample.
In order to illustrate the effectiveness of the method, fig. 4 compares the detection results of the method of the present invention with the detection results of the conventional CA-CFAR (constant false alarm rate) method, and it can be seen that the detection performance of the method of the present invention is better under the same false alarm rate.
It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Various modifications and alterations to this invention will become apparent to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims (5)
1. A method for detecting targets under a sea clutter background based on feature extraction is characterized by comprising the following steps:
s1, sampling in a sliding mode from an echo data starting point in a pulse direction by using a sliding window method, and storing a vector after each sliding window as an original data sample vector;
s2, carrying out time-frequency transformation on the original data sample vector,
s3, extracting a low-frequency eigenmode function according to the energy distribution characteristic of the time-frequency transformation result; the step S3 specifically includes:
s31, calculating the ratio of the signal energy of each intrinsic mode function to the signal energy of all the intrinsic mode functions;
s32, calculating the difference of the ratios corresponding to the adjacent intrinsic mode function signals, and recording the position of the maximum value of the ratio difference, so as to obtain a low-frequency intrinsic mode function;
s4, constructing a new data sample vector according to the extracted low-frequency eigenmode function;
s5, dividing the new data sample vector into a training set and a testing set;
s6, training the support vector machine by using a training set to obtain a weight vector and an offset of an optimal hyperplane;
s7, testing the weight vector of the optimal hyperplane and the support vector machine under bias by using a test set to obtain a target detection result; step S7 adopts an improved support vector machine, and the mathematical description thereof is as follows:
s.t.ξ i ≥0i＝1,2,...N
y i [k(ω,F i )-b]≥1-ξ i i＝1,2,...N
wherein ξ i For the relaxation variables, ω and b represent the weight vector and the offset, β, respectively, of the optimal hyperplane 0 And beta 1 Respectively representing penalty factors, k (omega, F), for the sea clutter samples and for the target samples 1 ) Is a Gaussian kernel function, F i Represents one sample, N represents the number of samples, y i = +1 denotes the SVM decision sample F i The result is the target sample, y i = -1 denotes SVM decision sample F i The result is a sea clutter sample.
2. The method for detecting the target under the sea clutter background based on the feature extraction as claimed in claim 1, wherein the width of the sliding window in step S1 is larger than the step length of the sliding window.
3. The method for detecting the target under the sea clutter background based on the feature extraction as claimed in claim 1 or 2, wherein the step S2 specifically comprises: and carrying out binary empirical mode decomposition on the original data sample vector stored after each sliding window to obtain a plurality of intrinsic mode functions.
4. The method for detecting the target under the sea clutter background based on the feature extraction as claimed in claim 3, wherein the step S4 specifically comprises: and calculating the energy of each low-frequency eigenmode function and adding the energy to obtain a new data sample vector.
5. The method for detecting the target under the sea clutter background based on the feature extraction as claimed in claim 1, wherein the process of obtaining the weight vector and the bias of the optimal hyperplane in the step S6 is as follows:
firstly, the false alarm probability P to be achieved is set fa A threshold psi, a penalty factor beta 0 And beta 1 、β 0 Is a value range boundary beta l And beta r ；
Utilizing the hyperplane determined by omega and b to classify the target and clutter samples and calculate the current false alarm probability P F ；
If P F And the false alarm probability P to be achieved fa Satisfy | P F -P fa |>Psi, then adjust beta according to the following two conditions 0 And beta 1 And continue the iteration:
if P is F <P fa Then set beta r ＝β 0 And beta is 0 ＝(β r +β l ) 2, continuing to iterate;
if P is F >P fa Then, set up β l ＝β 0 And beta is 0 ＝(β r +β l ) 2, continuing to iterate;
after each iteration, a new set of ω and b is obtained until | P F -P fa Stopping iteration when | is less than or equal to psi, and the omega and b at the moment are parameters of the optimal hyperplane.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210229272.3A CN114609602B (en) | 2022-03-09 | 2022-03-09 | Feature extraction-based target detection method under sea clutter background |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210229272.3A CN114609602B (en) | 2022-03-09 | 2022-03-09 | Feature extraction-based target detection method under sea clutter background |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114609602A CN114609602A (en) | 2022-06-10 |
| CN114609602B true CN114609602B (en) | 2023-04-07 |
Family
ID=81861930
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210229272.3A Active CN114609602B (en) | 2022-03-09 | 2022-03-09 | Feature extraction-based target detection method under sea clutter background |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114609602B (en) |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015054666A1 (en) * | 2013-10-10 | 2015-04-16 | Board Of Regents, The University Of Texas System | Systems and methods for quantitative analysis of histopathology images using multi-classifier ensemble schemes |
| CN104765979B (en) * | 2015-04-28 | 2018-04-24 | 南京信息工程大学 | A kind of sea clutter denoising method based on integrated empirical mode decomposition |
| US11169190B2 (en) * | 2016-09-19 | 2021-11-09 | The University Of New Hampshire | Techniques for empirical mode decomposition (EMD)-based noise estimation |
| CN106604229B (en) * | 2016-12-27 | 2020-02-18 | 东南大学 | Indoor positioning method based on manifold learning and improved support vector machine |
| CN106778694A (en) * | 2017-01-18 | 2017-05-31 | 北京工业大学 | A kind of gear transmission noises analysis method based on set empirical mode decomposition and SVMs |
| CN109934050A (en) * | 2017-12-15 | 2019-06-25 | 西安艾索信息技术有限公司 | A kind of improved ocean clutter cancellation algorithm decomposed based on EMD |
| CN108875799A (en) * | 2018-05-29 | 2018-11-23 | 齐鲁工业大学 | A kind of Mental imagery classifying identification method based on improvement S-transformation |
| CN108985159A (en) * | 2018-06-08 | 2018-12-11 | 平安科技（深圳）有限公司 | Human-eye model training method, eye recognition method, apparatus, equipment and medium |
| CN111985426A (en) * | 2020-08-27 | 2020-11-24 | 南京信息工程大学 | Sea clutter hybrid denoising algorithm based on variational modal decomposition |
-
2022
- 2022-03-09 CN CN202210229272.3A patent/CN114609602B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114609602A (en) | 2022-06-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106778610B (en) | Intra-pulse modulation identification method based on time-frequency image characteristics | |
| Shao et al. | Deep fusion for radar jamming signal classification based on CNN | |
| CN104569948B (en) | Sub-band adaptive GLRT LTD detection methods under sea clutter background | |
| CN103454624B (en) | The direct data domain moving target detection method of spectrum time empty based on dimensionality reduction sparse reconstruct | |
| CN103885057B (en) | Adaptive strain sliding window multi-object tracking method | |
| CN111610501B (en) | Method for detecting small target of sea radar | |
| CN109100696B (en) | Slow-speed moving target elimination method based on dot clutter map | |
| CN111505643B (en) | Sea surface small target detection method based on time-frequency image deep learning | |
| KR102275960B1 (en) | System and method for searching radar targets based on deep learning | |
| Wen et al. | Multiframe detection of sea-surface small target using deep convolutional neural network | |
| CN112881982A (en) | Method for restraining dense forwarding interference by frequency agile radar | |
| CN108387880A (en) | Detection method of small target under a kind of sea clutter background based on multiple dimensioned oriented Lyapunov indexes | |
| CN114609602B (en) | Feature extraction-based target detection method under sea clutter background | |
| CN110275150B (en) | Variable acceleration moving target coherent accumulation method based on empirical mode decomposition and iterative endpoint fitting | |
| Xiao et al. | Multi-target ISAR imaging based on image segmentation and short-time Fourier transform | |
| Li et al. | A bi-sru neural network based on soft attention for hrrp target recognition | |
| Lu et al. | Ship target recognition using high resolution range profiles based on FMT and SVM | |
| Pan et al. | GAN-CRT: A novel range-doppler estimation method in automotive radar systems | |
| Chen et al. | Integrated processing of radar detection and classification for moving target via time-frequency graph and CNN learning | |
| Gong et al. | Lightcnn: A compact cnn for moving maritime targets detection | |
| Cheng et al. | Time-frequency Feature-based Detection of Small Target on Sea-Surface | |
| Kenyon et al. | Range-Doppler ambiguity mitigation via closed-loop, adaptive PRF selection | |
| Cao et al. | An approach for high resolution radar target recognition based on BP neural network | |
| CN113030892B (en) | Sea surface small target detection method based on high-dimensional feature domain gradient lifting tree | |
| Xu et al. | Sea-Surface Floating Small Target Detection Based on SVM and Multidimensional Features |
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 |