CN113887583A - Radar RD image target detection method based on deep learning under low signal-to-noise ratio - Google Patents
Radar RD image target detection method based on deep learning under low signal-to-noise ratio Download PDFInfo
- Publication number
- CN113887583A CN113887583A CN202111083283.7A CN202111083283A CN113887583A CN 113887583 A CN113887583 A CN 113887583A CN 202111083283 A CN202111083283 A CN 202111083283A CN 113887583 A CN113887583 A CN 113887583A
- Authority
- CN
- China
- Prior art keywords
- radar
- target
- image
- neural network
- noise ratio
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 65
- 238000013135 deep learning Methods 0.000 title claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 39
- 238000013528 artificial neural network Methods 0.000 claims abstract description 33
- 238000000034 method Methods 0.000 claims abstract description 17
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 238000002372 labelling Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 238000012795 verification Methods 0.000 claims abstract description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 12
- 238000011176 pooling Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 11
- 238000003062 neural network model Methods 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 5
- 239000013598 vector Substances 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 239000003550 marker Substances 0.000 claims description 3
- 238000004088 simulation Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000003190 augmentative effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a radar RD image target detection method based on deep learning under low signal-to-noise ratio, which comprises the following steps: acquiring radar echo data and generating a radar RD image; preprocessing a radar RD image and labeling to obtain a data set; classifying the data set to obtain a training set, a verification set and a test set; constructing a deep learning neural network aiming at target detection under low signal-to-noise ratio; training the constructed neural network by adopting a training set, and outputting a loss value and a trained detection neural network; carrying out target detection on the test set by using the trained detection neural network; and obtaining a target detection accuracy result of the test set. The method is based on a large amount of radar RD image data, an optimal radar target detection network is obtained through deep learning neural network training, the target detection network obtained under the method can effectively detect the target of the radar RD image under the low signal to noise ratio, and the method has the advantages of high accuracy, good practical effect and the like.
Description
Technical Field
The invention relates to the technical field of target detection, in particular to a Range-Doppler (RD) image target detection method based on deep learning and under a low signal-to-noise ratio.
Background
In recent years, object detection has been widely used in many fields. The radar is an important means for detecting a target, and may analyze and process an echo in an irradiation region, detect target information from signals such as clutter, interference, and noise, and determine parameters such as a distance, a speed, and an angle of the target information. However, under a complex background, especially under a low signal-to-noise ratio, echo information often contains a large amount of noise and clutter information besides a target, the noise is close to the target amplitude, and the clutter often presents the characteristics of nonlinearity, non-gaussian, non-uniformity and non-stationarity, so that the radar target detection performance is greatly limited.
The existing radar target detection method comprises constant false alarm detection, a machine learning algorithm and the like, wherein the constant false alarm algorithm is based on a statistical model, is often difficult to accurately describe a background model, and can cause serious constant false alarm loss and lower detection performance under a non-uniform background, especially under low signal-to-noise ratios with different types and variable forms; target features are difficult to deeply extract by a machine learning algorithm, such as a target detection algorithm of a self-correcting extreme learning machine and the like, the capability of distinguishing a target from a background is weak, accurate classification of the target is difficult to realize, the difficulty of target detection is greatly increased, and the detection performance is limited. In conclusion, the existing radar RD image target detection method has the problems of simple model, low universality, weak learning ability and the like, and the problem of weak radar RD image target detection ability under low signal-to-noise ratio is difficult to fundamentally solve.
Disclosure of Invention
The invention discloses a radar RD image target detection method under a low signal-to-noise ratio based on deep learning, which can carry out deep extraction on target characteristics under a low signal-to-noise ratio background through a neural network, greatly distinguish a target from the background and further improve the detection accuracy.
The technical solution for realizing the purpose of the invention is as follows: a radar RD image target detection method under low signal-to-noise ratio based on deep learning comprises the following steps:
step 1, radar echo data are obtained, and a radar RD image is generated;
step 3, classifying the data set to obtain a training set, a verification set and a test set;
step 4, constructing a deep learning neural network aiming at target detection under low signal-to-noise ratio;
step 5, training the constructed neural network by adopting a training set, and outputting a loss value and a trained detection neural network;
step 6, carrying out target detection on the test set by using the trained detection neural network;
and 7, obtaining a target detection accuracy result of the test set.
Further, the step 1 of acquiring radar echo data and generating a radar RD image further comprises,
step 1.1, adding random noises with different signal-to-noise ratios into the generated radar echo data through simulation;
and 1.2, randomly generating the number, the position and the speed of targets in the radar echo data within a set range.
Further, the step 2 of preprocessing the radar RD image and labeling specifically includes:
step 2.1, standardizing the radar RD image, and adjusting the resolution to be matched with deep learning neural network learning;
and 2.2, acquiring a specific coordinate value of the position of the target, and setting a label according to the coordinate value of the position of the target to generate standard label data.
Further, the training set of step 3 includes a training image data set and a label data set.
Further, the step 4 of constructing a deep learning neural network for target detection at a low signal-to-noise ratio is performed, wherein:
the low signal-to-noise ratio is defined as the signal-to-noise ratio lower than 7dB after pulse boosting and fast Fourier transform;
the deep learning neural network adopts an Faster R-CNN neural network model and comprises a convolutional layer, a regional suggestion network, a pooling layer and a full-connection layer; after image data is input, firstly, extracting deep features of an image through a convolutional layer, performing convolution operation on the input image by using VGG16, and reducing the dimension of the image by using a maximum pooling layer after the convolutional layer to generate a feature map, wherein the feature map comprises deep feature information of the image; inputting the feature map into a region suggestion network, extracting features through a sliding window to obtain a plurality of suggestion regions, inputting the feature map with the suggestion regions into an interested region pooling layer, and outputting the feature map with a constant size, namely changing target suggestion regions with different sizes into feature vectors with the same size for output; finally, carrying out classification and regression operation through a full connection layer, wherein the classification layer is used for distinguishing the target and the complex background and constructing the maximum difference between the target and the complex background, and the regression layer is used for calculating the distance between the target position and the real position; the classification layer inputs feature information with a fixed size, extracts the region category through the full connection layer and the softmax and outputs the confidence coefficient of the region category, the regression layer obtains the position deviation of each target suggestion region through frame regression, generates a detection frame closer to the real position of the target, and finally obtains the final accurate position of the optimal region category and the boundary frame regression detection frame.
Further, the step 5 of training the constructed neural network by using the training set comprises the following steps,
step 5.1, respectively training the radar range-Doppler image data sets under different signal-to-noise ratios;
step 5.2, inputting the training image data set and the label data set into the established Faster R-CNN neural network model, and calculating a loss value;
step 5.3, iteratively optimizing the fast R-CNN neural network model parameters, and repeatedly training;
and 5.4, finishing training and outputting a detection neural network when the loss value reaches the optimal value.
Further, the loss value in step 5 is calculated by a loss function L (-) according to the following formula:
wherein p isiIn order to predict the probability of a classification,
for true classification, tiIn order to be able to predict the parametric coordinates,
for true parametric coordinates, NclsIs the minimum batch size, NregIs the number of marker boxes, λ is the weight balance parameter, L is the total loss function, LclsTwo classes of logarithmic loss, LregIs the regression loss.
Further, the accuracy of step 7 includes a detection rate PdAnd false alarm rate PfThe calculation formula is as follows,
wherein TP is a true example, and the true target is predicted to be the number of targets; FN is false negative example, and the real target is predicted to be the number of non-targets; FP is a false positive case and non-target predictions are the number of targets.
Compared with the prior art, the invention has the following remarkable advantages: (1) based on radar RD image data under a large number of different signal-to-noise ratios, the method has universality; (2) the radar RD image detection precision under low signal-to-noise ratio is improved through fast R-CNN algorithm detection; (3) and a training loss function is optimized, the detection precision is improved, and the method is suitable for popularization and application in practical application.
Drawings
FIG. 1 is a schematic flow chart of a method for detecting a radar RD image target under a low signal-to-noise ratio based on deep learning according to the present invention.
Fig. 2 is a graph of RD at low signal-to-noise ratio in the present invention.
Fig. 3 is a plot of the RD labeled in the present invention at low signal-to-noise ratio.
FIG. 4 is a diagram of the basic structure of the object detection algorithm of the present invention.
FIG. 5 is a graph of a model loss function in the present invention.
Fig. 6 is a graph of the detection of an RD image at low signal-to-noise ratio according to the present invention.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
The method comprises the steps of firstly establishing a radar RD image data set under different signal-to-noise ratios, ensuring the randomness of the data set, carrying out data annotation to obtain a training set, a verification set and a test set, training the RD image data set by using a Faster R-CNN network, and sending extracted features into a neural network to obtain feature vectors with screening and weighting characteristics. For important target features, the network can assign larger processing weight, thereby enhancing the feature learning capability of the network on the target area. And finally, optimizing a loss function, and realizing high-precision detection of the radar RD image under the low signal-to-noise ratio.
With reference to fig. 1 to 6, the present invention provides a method for detecting a target in a radar RD image under a low signal-to-noise ratio based on deep learning, which includes the following steps:
step 1, radar echo data are obtained, and a radar RD image is generated;
step 3, classifying the data set to obtain a training set, a verification set and a test set;
step 4, constructing a deep learning neural network aiming at target detection under low signal-to-noise ratio;
step 5, training the constructed neural network by adopting a training set, and outputting a loss value and a trained detection neural network;
step 6, carrying out target detection on the test set by using the trained detection neural network;
and 7, obtaining a target detection accuracy result of the test set.
In a specific embodiment, the step 1 of acquiring radar echo data and generating a radar range-doppler dimensional image further comprises,
step 1.1, adding random noises with different signal-to-noise ratios into the generated radar echo data through simulation;
and 1.2, randomly generating the number, the position and the speed of targets in the radar echo data within a set range.
As a specific embodiment, the preprocessing and labeling of the original image in step 2 includes,
step 2.1, standardizing the radar RD image, and adjusting the resolution to be matched with deep learning neural network learning;
and 2.2, acquiring a specific coordinate value of the position of the target, and setting a label according to the coordinate value of the position of the target to generate standard label data.
In a specific embodiment, the training set in step 3 includes a training image data set and a label data set.
As a specific embodiment, the deep learning neural network for target detection at low signal-to-noise ratio is constructed in step 4, wherein:
the low signal-to-noise ratio is defined as the signal-to-noise ratio lower than 7dB after pulse boosting and fast Fourier transform;
the deep learning neural network adopts an Faster R-CNN neural network model and comprises a convolutional layer, a regional suggestion network, a pooling layer and a full-connection layer; after image data is input, firstly, extracting deep features of an image through a convolutional layer, performing convolution operation on the input image by using VGG16, and reducing the dimension of the image by using a maximum pooling layer after the convolutional layer to generate a feature map, wherein the feature map comprises deep feature information of the image; inputting the feature map into a region suggestion network, extracting features through a sliding window to obtain a plurality of suggestion regions, inputting the feature map with the suggestion regions into an interested region pooling layer, and outputting the feature map with a constant size, namely changing target suggestion regions with different sizes into feature vectors with the same size for output; finally, carrying out classification and regression operation through a full connection layer, wherein the classification layer is used for distinguishing the target and the complex background and constructing the maximum difference between the target and the complex background, and the regression layer is used for calculating the distance between the target position and the real position; the classification layer inputs feature information with a fixed size, extracts the region category through the full connection layer and the softmax and outputs the confidence coefficient of the region category, the regression layer obtains the position deviation of each target suggestion region through frame regression, generates a detection frame closer to the real position of the target, and finally obtains the final accurate position of the optimal region category and the boundary frame regression detection frame.
In a specific embodiment, the training of step 5 further comprises the following steps,
step 5.1, respectively training the radar range-Doppler image data sets under different signal-to-noise ratios;
step 5.2, inputting the training image data set and the label data set into the established Faster R-CNN neural network model, and calculating a loss value;
step 5.3, iteratively optimizing the fast R-CNN neural network model parameters, and repeatedly training;
and 5.4, finishing training and outputting a detection neural network when the loss value reaches the optimal value.
As a specific implementation manner, the loss value in step 5 is calculated by a loss function L (-) according to the following formula:
wherein p isiIn order to predict the probability of a classification,
for true classification, tiIn order to be able to predict the parametric coordinates,
for true parametric coordinates, NclsTo the minimum batch size，NregIs the number of marker boxes, λ is the weight balance parameter, L is the total loss function, LclsTwo classes of logarithmic loss, LregIs the regression loss.
As a specific embodiment, the accuracy in step 7 includes a detection rate PdAnd false alarm rate PfThe calculation formula is as follows,
wherein TP is a true example, and the true target is predicted to be the number of targets; FN is false negative example, and the real target is predicted to be the number of non-targets; FP is a false positive case and non-target predictions are the number of targets.
The invention is described in further detail below with reference to the figures and the embodiments.
Examples
The embodiment provides a method for detecting a radar RD image target under a low signal-to-noise ratio based on deep learning, data used for an experiment are simulated radar RD images, target information is included, the number, the position, the speed and the like of targets are included, and target parameters are ensured to be random within a certain range; the radar echo data also comprises random noises with different signal-to-noise ratios, and a certain amount of low signal-to-noise ratio data exists; the original echo data is converted into an image in a range dimension and a doppler dimension, and it is ensured that a target and interference (noise, clutter, etc.) exist in the image at the same time, wherein the target occupies one pixel, the resolution of the generated RD image is 256 × 225, the range dimension is 225 cells, and the doppler dimension is 256 cells, as shown in fig. 2.
The RD image is first pre-processed and the target is marked, with the results shown in fig. 3. And then, sending the augmented data set into a Faster R-CNN network for feature extraction and feature fusion, wherein the network structure is as shown in FIG. 4, the VGG16 is used for performing convolution operation on the input image, and the deep convolution network model has 16 layers: 13 convolutional layers and 3 fully connected layers. After the convolutional layers, the dimensionality of the images is reduced using the maximum pooling layer, and the images are classified by SoftMax, generating a feature map. The area suggestion network inputs the feature map provided by the convolutional layer and extracts the features through a sliding window. SoftMax determines whether the anchor point located at the center of the sliding window is a positive or negative value. Nine regions are generated because there are three different sizes: 128. 256 and 512, and three different ratios: 1, 2, 1, including information on the location and size of the region, and sent to the pooling layer along with the signature maps obtained from the convolutional layers. The pooling layer converts the different sized inputs into fixed lengths, converts the target region or clutter region into vectors of the same size, and sends them to the subsequent fully-connected layer. And finally, calculating the region category and the boundary box regression to obtain the final accurate position of the detection box.
And finally, predicting the target position, and using the optimized loss function when the model parameters are updated reversely, wherein the result of the loss function is shown in fig. 5, and the detection result of the method is shown in fig. 6.
Claims (8)
1. A radar RD image target detection method under low signal-to-noise ratio based on deep learning is characterized by comprising the following steps:
step 1, radar echo data are obtained, and a radar RD image is generated;
step 2, preprocessing the radar RD image and labeling to obtain a data set;
step 3, classifying the data set to obtain a training set, a verification set and a test set;
step 4, constructing a deep learning neural network aiming at target detection under low signal-to-noise ratio;
step 5, training the constructed neural network by adopting a training set, and outputting a loss value and a trained detection neural network;
step 6, carrying out target detection on the test set by using the trained detection neural network;
and 7, obtaining a target detection accuracy result of the test set.
2. The method for detecting the target of the radar RD image under the low signal-to-noise ratio based on the deep learning as claimed in claim 1, wherein the step 1 of acquiring the radar echo data and generating the radar RD image further comprises,
step 1.1, adding random noises with different signal-to-noise ratios into the generated radar echo data through simulation;
and 1.2, randomly generating the number, the position and the speed of targets in the radar echo data within a set range.
3. The method for detecting the radar RD image target under the low signal-to-noise ratio based on the deep learning as claimed in claim 1, wherein the preprocessing and labeling the radar RD image in the step 2 specifically comprises:
step 2.1, standardizing the radar RD image, and adjusting the resolution to be matched with deep learning neural network learning;
and 2.2, acquiring a specific coordinate value of the position of the target, and setting a label according to the coordinate value of the position of the target to generate standard label data.
4. The deep learning-based radar RD image target detection method under low signal-to-noise ratio as claimed in claim 1, wherein the training set of step 3 comprises a training image data set and a label data set.
5. The method for detecting the target of the radar RD image under the low signal-to-noise ratio based on the deep learning as claimed in claim 1, wherein the deep learning neural network for the target detection under the low signal-to-noise ratio is constructed in step 4, wherein:
the low signal-to-noise ratio is defined as the signal-to-noise ratio lower than 7dB after pulse boosting and fast Fourier transform;
the deep learning neural network adopts an Faster R-CNN neural network model and comprises a convolutional layer, a regional suggestion network, a pooling layer and a full-connection layer; after image data is input, firstly, extracting deep features of an image through a convolutional layer, performing convolution operation on the input image by using VGG16, and reducing the dimension of the image by using a maximum pooling layer after the convolutional layer to generate a feature map, wherein the feature map comprises deep feature information of the image; inputting the feature map into a region suggestion network, extracting features through a sliding window to obtain a plurality of suggestion regions, inputting the feature map with the suggestion regions into an interested region pooling layer, and outputting the feature map with a constant size, namely changing target suggestion regions with different sizes into feature vectors with the same size for output; finally, carrying out classification and regression operation through a full connection layer, wherein the classification layer is used for distinguishing the target and the complex background and constructing the maximum difference between the target and the complex background, and the regression layer is used for calculating the distance between the target position and the real position; the classification layer inputs feature information with a fixed size, extracts the region category through the full connection layer and the softmax and outputs the confidence coefficient of the region category, the regression layer obtains the position deviation of each target suggestion region through frame regression, generates a detection frame closer to the real position of the target, and finally obtains the final accurate position of the optimal region category and the boundary frame regression detection frame.
6. The method for detecting the radar RD image target under the low signal-to-noise ratio based on the deep learning as claimed in claim 5, wherein the step 5 of training the constructed neural network by using the training set comprises the following steps,
step 5.1, respectively training the radar range-Doppler image data sets under different signal-to-noise ratios;
step 5.2, inputting the training image data set and the label data set into the established Faster R-CNN neural network model, and calculating a loss value;
step 5.3, iteratively optimizing the fast R-CNN neural network model parameters, and repeatedly training;
and 5.4, finishing training and outputting a detection neural network when the loss value reaches the optimal value.
7. The method for detecting the radar RD image target under the low signal-to-noise ratio based on the deep learning as claimed in claim 1, wherein the loss value in the step 5 is calculated by a loss function L (-) according to the following formula:
wherein p isiIn order to predict the probability of a classification,
for true classification, tiIn order to be able to predict the parametric coordinates,
for true parametric coordinates, NclsIs the minimum batch size, NregIs the number of marker boxes, λ is the weight balance parameter, L is the total loss function, LclsTwo classes of logarithmic loss, LregIs the regression loss.
8. The method as claimed in claim 1, wherein the accuracy of step 7 includes a detection rate PdAnd false alarm rate PfThe calculation formula is as follows,
wherein TP is a true example, and the true target is predicted to be the number of targets; FN is false negative example, and the real target is predicted to be the number of non-targets; FP is a false positive case and non-target predictions are the number of targets.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111083283.7A CN113887583A (en) | 2021-09-16 | 2021-09-16 | Radar RD image target detection method based on deep learning under low signal-to-noise ratio |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111083283.7A CN113887583A (en) | 2021-09-16 | 2021-09-16 | Radar RD image target detection method based on deep learning under low signal-to-noise ratio |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113887583A true CN113887583A (en) | 2022-01-04 |
Family
ID=79009605
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111083283.7A Pending CN113887583A (en) | 2021-09-16 | 2021-09-16 | Radar RD image target detection method based on deep learning under low signal-to-noise ratio |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113887583A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114994674A (en) * | 2022-08-08 | 2022-09-02 | 中国科学技术大学 | Intelligent microwave staring correlated imaging method, equipment and storage medium |
| CN117452367A (en) * | 2023-12-21 | 2024-01-26 | 西安电子科技大学 | SAR load radiation signal extraction method and device based on broadband imaging radar |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110824450A (en) * | 2019-10-15 | 2020-02-21 | 中国人民解放军国防科技大学 | Radar target HRRP robust identification method in noise environment |
| CN111722199A (en) * | 2020-08-10 | 2020-09-29 | 上海航天电子通讯设备研究所 | Radar signal detection method based on convolutional neural network |
| 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 |
-
2021
- 2021-09-16 CN CN202111083283.7A patent/CN113887583A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110824450A (en) * | 2019-10-15 | 2020-02-21 | 中国人民解放军国防科技大学 | Radar target HRRP robust identification method in noise environment |
| CN111722199A (en) * | 2020-08-10 | 2020-09-29 | 上海航天电子通讯设备研究所 | Radar signal detection method based on convolutional neural network |
| 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 |
Non-Patent Citations (1)
| Title |
|---|
| 张暄等: "基于卷积神经网络的雷达目标检测方法", 计算机测量与控制, 28 February 2021 (2021-02-28), pages 49 - 52 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114994674A (en) * | 2022-08-08 | 2022-09-02 | 中国科学技术大学 | Intelligent microwave staring correlated imaging method, equipment and storage medium |
| CN114994674B (en) * | 2022-08-08 | 2022-12-23 | 中国科学技术大学 | Intelligent microwave staring correlated imaging method and equipment and storage medium |
| CN117452367A (en) * | 2023-12-21 | 2024-01-26 | 西安电子科技大学 | SAR load radiation signal extraction method and device based on broadband imaging radar |
| CN117452367B (en) * | 2023-12-21 | 2024-03-26 | 西安电子科技大学 | SAR load radiation signal extraction method and device based on broadband imaging radar |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516561B (en) | SAR image target identification method based on DCGAN and CNN | |
| CN110210463B (en) | Precise ROI-fast R-CNN-based radar target image detection method | |
| Lei et al. | Automatic hyperbola detection and fitting in GPR B-scan image | |
| CN110472627B (en) | End-to-end SAR image recognition method, device and storage medium | |
| CN107229918B (en) | SAR image target detection method based on full convolution neural network | |
| CN112132042B (en) | SAR image target detection method based on contrast domain adaptation | |
| CN113486961A (en) | Radar RD image target detection method and system based on deep learning under low signal-to-noise ratio and computer equipment | |
| CN108447057B (en) | SAR image change detection method based on significance and depth convolution network | |
| CN107808138B (en) | Communication signal identification method based on FasterR-CNN | |
| CN108168564A (en) | A kind of Data Association based on LHD grey relational grades | |
| CN113887583A (en) | Radar RD image target detection method based on deep learning under low signal-to-noise ratio | |
| Nuhoglu et al. | Image segmentation for radar signal deinterleaving using deep learning | |
| CN111401168A (en) | Multi-layer radar feature extraction and selection method for unmanned aerial vehicle | |
| CN115937659A (en) | Mask-RCNN-based multi-target detection method in indoor complex environment | |
| CN114973300B (en) | Component type identification method and device, electronic equipment and storage medium | |
| CN111368653B (en) | Low-altitude small target detection method based on R-D graph and deep neural network | |
| CN113850838A (en) | Ship voyage intention acquisition method and device, computer equipment and storage medium | |
| Sun et al. | Ship detection in SAR images based on multilevel superpixel segmentation and fuzzy fusion | |
| CN116469045A (en) | Image recognition matching method and system based on deep learning | |
| CN113361639A (en) | Deep learning-based radiation source signal multi-model comprehensive classification method | |
| CN113989632A (en) | Bridge detection method and device for remote sensing image, electronic equipment and storage medium | |
| CN117111013B (en) | Radar target tracking track starting method, device, equipment and medium | |
| CN114283296B (en) | Target identification self-evaluation method based on structural similarity | |
| Yu et al. | Mine like object detection and recognition based on intrackability and improved BOW | |
| Pan et al. | An effective marine target detection algorithm based on YOLOv3 |
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 |