CN108664933B - Training method of convolutional neural network for SAR image ship classification, classification method of convolutional neural network and ship classification model - Google Patents
Training method of convolutional neural network for SAR image ship classification, classification method of convolutional neural network and ship classification model Download PDFInfo
- Publication number
- CN108664933B CN108664933B CN201810450109.3A CN201810450109A CN108664933B CN 108664933 B CN108664933 B CN 108664933B CN 201810450109 A CN201810450109 A CN 201810450109A CN 108664933 B CN108664933 B CN 108664933B
- Authority
- CN
- China
- Prior art keywords
- training
- ship
- neural network
- convolutional neural
- sar image
- 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
- 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
- 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
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
Abstract
The invention provides a training method of a convolutional neural network for SAR image ship classification and a classification method thereof, wherein the training method comprises the following steps: acquiring a slice with a ship category in an SAR image; training a convolutional neural network for SAR image ship classification based on the slice with the ship category so as to enable the convolutional neural network to reach preset training precision; the convolution neural network for SAR image ship classification is constructed by removing a first full-connection layer of a first model on the basis of the first model and adding a second full-connection layer according to the number of ship types of the slices with the ship types. The training method can be used for training the convolutional neural network for SAR image ship classification under the condition that only a small amount of training data exists, so that the preset training precision is achieved. And the convolutional neural network for SAR image ship classification, which reaches the preset training precision, is applied to SAR image ship classification, and the ship classification precision can reach 97.54%.
Description
Technical Field
The application relates to the field of remote sensing, in particular to a training method of a convolutional neural network for SAR image ship classification and a classification method thereof.
Background
Since 2007, there are many high-resolution SAR satellites that have been successfully transmitted, such as Cosmo-SkyMed, TerraSAR-X, ALOS-2PALSAR-2, Gaofen-3, etc., and the high-resolution SAR images obtained based on satellites have a resolution of over 3 meters, which contains abundant information about ground features, such as the geometric characteristics of ships, making it possible to distinguish between different classes of ships.
Deep learning models (e.g., convolutional neural networks) save time in feature extraction and selection and in optimizing classifiers by being able to automatically learn information representing surface features in SAR images and provide corresponding end-to-end processing without human intervention. Therefore, the deep learning model is gradually applied to the SAR image for classification task. However, the bottleneck in using a deep learning model is that it requires a large amount of training data, and obtaining a large amount of label data is time consuming and difficult to obtain. Therefore, a new method for training a deep learning model is needed.
Disclosure of Invention
The invention provides a training method of a convolutional neural network for SAR image ship classification and a classification method thereof. The training method can be used for training the convolutional neural network for SAR image ship classification under the condition that only a small amount of training data exists, so that the preset training precision is achieved. And the convolutional neural network for SAR image ship classification, which reaches the preset training precision, is applied to SAR image ship classification, and the ship classification precision can reach 97.54%.
In order to solve the above technical problem, an embodiment of the present invention provides the following technical solutions:
the invention provides a training method of a convolutional neural network for SAR image ship classification, which comprises the following steps:
acquiring a slice with a ship category in an SAR image;
training a convolutional neural network for SAR image ship classification based on the slice with the ship class so as to enable the convolutional neural network to reach preset training precision;
the convolutional neural network for SAR image ship classification is constructed by removing a first full-connection layer of a first model on the basis of the first model and adding a second full-connection layer according to the number of ship types of the slices with the ship types.
Preferably, the first fully connected layer comprises a first Softmax layer and two first fully connected layers.
Preferably, the second fully connected layer comprises a second Softmax layer and a second fully connected layer.
Preferably, the acquiring the slice with the ship category in the SAR image includes:
and acquiring the section with the ship category according to ship information and a preset rule on the basis of the SAR image.
Preferably, the training of the convolutional neural network for the ship classification of the SAR image based on the slice with the ship class includes training of all connection layers of the convolutional neural network for the ship classification of the SAR image, or
And training a second full connection layer of the convolutional neural network for SAR image ship classification.
Preferably, the training of the convolutional neural network for the ship classification of the SAR image based on the slice with the ship class includes training of all connection layers of the convolutional neural network for the ship classification of the SAR image, or
Training a second fully-connected layer and a second Softmax layer of the convolutional neural network for SAR image ship classification.
Preferably, training the convolutional neural network for the ship classification of the SAR image based on the slice with the ship class so that the convolutional neural network can reach a preset training precision, includes:
setting initialization parameters of convolutional neural network training for SAR image ship classification;
inputting the slices with the ship categories into a convolutional neural network for SAR image ship classification to be trained to obtain an output result;
judging whether the output result reaches the preset training precision:
if not, adjusting the initialization parameters, and continuing to execute the steps until the output result reaches the preset training precision.
Preferably, the initialization parameters for convolutional neural network training for SAR image ship classification are set, including,
and pre-training parameters obtained by pre-training the first model are used as initialization parameters of the convolutional neural network for SAR image ship classification.
Preferably, the first model is pre-trained on the optical data set to obtain pre-training parameters.
Preferably, the first model is selected from VGG16, VGG19, inclusion v3 or Xception.
The second aspect of the present invention provides a method for ship classification of an SAR image based on a convolutional neural network obtained after training by the above training method for a convolutional neural network for ship classification of an SAR image, which includes:
acquiring a ship slice in the SAR image to be classified;
and inputting the ship slice into a convolutional neural network which reaches the preset training precision and is used for SAR image ship classification to obtain a ship classification result of the ship slice.
The invention provides a ship classification model, wherein the model is realized based on a convolutional neural network, and the convolutional neural network is constructed by removing a first full-link layer of the first model on the basis of the first model and adding a second full-link layer according to the number of ship types of the slices with the ship types.
Preferably, the first fully connected layer comprises a first Softmax layer and two first fully connected layers.
Preferably, the second fully connected layer comprises a second Softmax layer and a second fully connected layer.
Preferably, the first model is selected from VGG16, VGG19, inclusion v3 or Xception.
Based on the disclosure of the above embodiments, it can be known that the embodiments of the present invention have the following beneficial effects: the training method of the convolutional neural network for SAR image ship classification provided by the invention can realize training of the convolutional neural network for SAR image ship classification under the condition that only a small amount of training data exists, namely, a first model is pre-trained on an optical data set to obtain pre-training parameters; the pre-training parameters are used as initialization parameters for training the convolutional neural network for SAR image ship classification instead of carrying out random assignment, and then the convolutional neural network for SAR image ship classification is trained by adopting two types of methods to achieve the preset training precision. Experiments show that the verification precision obtained by the first training method is 2% higher than that obtained by the second training method, which indicates that the first training method is more suitable for training the convolutional neural network for SAR image ship classification in the invention.
The convolutional neural network for SAR image ship classification, which achieves the preset training precision, is applied to SAR image ship classification, and experiments show that under the same conditions, the convolutional neural network based on VGG16 has higher ship classification precision than the convolutional neural networks based on other three models, and the ship classification precision can reach 97.54%.
Drawings
FIG. 1 is a flowchart of a convolutional neural network training method for SAR image ship classification according to an embodiment of the present invention;
FIG. 2 is a slice with vessel categories for three different categories according to an embodiment of the present invention;
FIG. 3 is a process diagram of pre-training the VGG16 model on the ImageNet dataset using Gaussian distribution initialization parameters according to an embodiment of the present invention;
fig. 4 is a process diagram of training a convolutional neural network based on the VGG16 model by two types of methods according to an embodiment of the present invention.
Detailed Description
The following detailed description of specific embodiments of the present invention is provided in connection with the accompanying drawings, which are not intended to limit the invention.
It will be understood that various modifications may be made to the embodiments disclosed herein. Accordingly, the foregoing description should not be construed as limiting, but merely as exemplifications of embodiments. Other modifications will occur to those skilled in the art within the scope and spirit of the disclosure.
The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the disclosure and, together with a general description of the disclosure given above, and the detailed description of the embodiments given below, serve to explain the principles of the disclosure.
These and other characteristics of the invention will become apparent from the following description of a preferred form of embodiment, given as a non-limiting example, with reference to the accompanying drawings.
It should also be understood that, although the invention has been described with reference to some specific examples, a person of skill in the art shall certainly be able to achieve many other equivalent forms of the invention, having the characteristics as set forth in the claims and hence all coming within the field of protection defined thereby.
The above and other aspects, features and advantages of the present disclosure will become more apparent in view of the following detailed description when taken in conjunction with the accompanying drawings.
Specific embodiments of the present disclosure are described hereinafter with reference to the accompanying drawings; however, it is to be understood that the disclosed embodiments are merely examples of the disclosure that may be embodied in various forms. Well-known and/or repeated functions and structures have not been described in detail so as not to obscure the present disclosure with unnecessary or unnecessary detail. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a basis for the claims and as a representative basis for teaching one skilled in the art to variously employ the present disclosure in virtually any appropriately detailed structure.
The specification may use the phrases "in one embodiment," "in another embodiment," "in yet another embodiment," or "in other embodiments," which may each refer to one or more of the same or different embodiments in accordance with the disclosure.
The experiment of the invention is carried out on a Ubuntu system, and the NVIDAI GPU GTX1070 video card with 8G memory is provided. And the experiment was performed on Keras (developed on the basis of tenserflow, CNKT and thano). The first model was downloaded from the Kearas official network.
The embodiments of the present invention will be described in detail below with reference to the accompanying drawings:
as shown in fig. 1, a first embodiment of the present application provides a training method of a convolutional neural network for SAR image ship classification, and the following describes the present embodiment in detail.
Step S101 acquires a slice with a ship category in the SAR image.
In the experiment, 6-scene SAR images acquired by a Cosmo-Skymed satellite are taken as research objects, and the specific description of the SAR images is shown in Table 1.
TABLE 1 SAR image information
And acquiring the section with the ship category according to the ship information and a preset rule on the basis of the SAR image.
In one embodiment of the invention, a 256 pixel x 256 pixel slice with a ship category is intercepted based on the SAR image described above, based on AIS information and expert guidance. Wherein 446 pieces of slices with ship categories are obtained, the ship categories mainly include cargo ships, container ships and oil tankers, and the specific number distribution of three different categories of ships is shown in table 2. Three different categories of slices with vessel categories are shown in figure 2.
TABLE 2 detailed number table of different classes of ships
Step S102 is to construct a convolutional neural network for SAR image ship classification.
The convolutional neural network for SAR image ship classification is constructed by removing a first full-connection layer of a first model on the basis of the first model and adding a second full-connection layer according to the number of ship classes of the slice with the ship classes.
Wherein the first fully connected layer comprises a first Softmax layer and two first fully connected layers. The output class of the first Softmax layer is 1000, and the number of neurons of the first fully connected layer is 4096.
Wherein the second fully connected layer comprises a second Softmax layer and a second fully connected layer. The output class of the second Softmax layer is 3, and the number of neurons of the second fully connected layer is 128.
In other embodiments of the present invention, the second fully connected layer comprises a second Softmax layer, a second full connected layer, and a first "Dropout" layer.
In another embodiment of the invention, the first model is selected from VGG16, VGG19, inclusion v3 or Xception. Specific information for the four different first models is shown in table 3.
TABLE 3 four different first model information
In one embodiment of the invention, the number of said ship classes may be more than 2, for example from 2 to 10. If the ship categories of the slices with the ship category are cargo ships, container ships, and tankers, the number of the ship categories at this time is 3.
As shown in fig. 3 and 4, the construction of the convolutional neural network with the VGG16 model as the first model is shown, and it can be seen from the figure that the largest difference between the convolutional neural network based on the VGG16 model and the VGG16 model is at the top layer, that is, there are two neuron numbers with output class 1000 of 4096 in the VGG16 model, and there are a second Softmax layer with output class 3 and a second fully connected layer with neuron number 128 in the convolutional neural network constructed based on the VGG16 model.
Step S103 trains a convolutional neural network for SAR image ship classification.
Step S103-1, acquiring initialization parameters of convolutional neural network training for SAR image ship classification:
because training data (referred to as slices with ship classes in the present invention) for training the convolutional neural network for ship classification of SAR images is limited, it is necessary to pre-train the first model on an optical data set (e.g., ImageNet or Coco) with a large amount of data, and to use the pre-trained parameters as initialization parameters for training the convolutional neural network for ship classification of SAR images, rather than randomly assigning values to the pre-trained parameters. Fig. 3 shows a process of pre-training ImageNet data set using gaussian distribution initialization parameters, taking VGG16 model as an example. Namely, the specific process for acquiring the initialization parameters of the convolutional neural network training for the SAR image ship classification is as follows:
pre-training the first model on an optical data set to obtain pre-training parameters;
and taking the pre-training parameters as initialization parameters for training the convolutional neural network for SAR image ship classification.
And step S103-2, inputting the section with the ship category obtained in the step S101 into a convolutional neural network for SAR image ship classification to train, and obtaining an output result.
In the above steps, although the pre-training parameters obtained by pre-training the first model on the optical data set are used as the initialization parameters for training the convolutional neural network for the SAR image ship classification, the pre-training parameters obtained by pre-training on the optical data set may not be suitable for the SAR image in consideration of the difference between the optical image and the SAR image, such as the difference between the imaging mechanism and the difference between information about the target, and therefore, the convolutional neural network for the SAR image ship classification still needs to be trained.
The invention adopts two methods to train the convolutional neural network for SAR image ship classification:
the first training method is as follows: training all the connecting layers;
the second type of training method is: only the second fully connected layer is trained.
In one embodiment of the invention, the convolutional neural network for the ship classification of the SAR image can be trained by two types of training methods, wherein the first type is that all connection layers of the convolutional neural network for the ship classification of the SAR image are trained; the second type is that only the second fully connected layer of the convolutional neural network for SAR image ship classification is trained.
As shown in fig. 4, the procedure of training the convolutional neural network based on the VGG16 model is shown as two types of methods, and as can be seen from the figure, the first type of method trains all the connected layers, while the second type of method trains only the "full connected" connected layer and the "Softmax" connected layer in the frame of "newly added network layer" shown in the figure.
In yet another embodiment of the present invention, the first class of training method is a training mode by random gradient descent with a learning rate of 0.0001 and moment of 0.99.
In other embodiments of the present invention, the second type of training method is a training mode with a learning rate of 0.001 and moment of 0.9.
And S103-3, training the convolutional neural network for SAR image ship classification to reach preset training precision.
Judging whether the output result reaches the preset training precision or not;
if not, adjusting the initialization parameters, and continuing to execute the steps until the output result reaches the preset training precision.
In one embodiment of the present invention, the condition for judging that the output result reaches the preset training accuracy is that the output result or the loss hardly changes.
And step S104, verifying the convolutional neural network which reaches the preset training precision and is used for SAR image ship classification to obtain verification precision.
The invention respectively adopts the two training methods (namely, the first training method is to train all the connection layers, and the second training method is to train only the second full connection layer) for 4 different convolutional neural networks for SAR image ship classification (the 4 different convolutional neural networks for SAR image ship classification are respectively called convolutional neural network based on Incepion V3, convolutional neural network based on VGG16, convolutional neural network based on VGG19 and convolutional neural network based on Xception), so as to obtain the preset training precision and the verification precision of each convolutional neural network under the two different training methods, and the specific experimental result is shown in Table 4.
TABLE 4 different precision tables corresponding to convolutional neural networks using different training methods
The experimental results show that, on one hand, the first type of training method is more stable than the second type of training method in process, and on the other hand, as can be seen from the table, the verification accuracy of the four different convolutional neural networks obtained by using the first type of training method is 2% higher than that obtained by using the second type of training method. The reason for the above phenomenon may be that the convolutional neural network obtained through training by the first training method is more suitable for the SAR image.
A second embodiment of the present application provides a method for ship classification of an SAR image based on a convolutional neural network obtained after training by the above training method for a convolutional neural network for ship classification of an SAR image, including:
acquiring a ship slice in the SAR image to be classified;
and inputting the ship slice into a convolutional neural network which reaches the preset training precision and is used for SAR image ship classification to obtain a ship classification result of the ship slice.
Experiments show that compared with the VGG16 model, although the Top-1 accuracy, the Top-5 accuracy and the model depth of other models are higher than those of the VGG16 model, the convolutional neural network based on the VGG16 obtains higher ship classification accuracy than other models, and the ship classification accuracy can reach 97.54%. The reason for this may be that the model with deeper model depths extracts features that are more and more abstract and thus more suitable for optical images and not for SAR images.
The third embodiment of the present application provides a ship classification model, the structure of which is:
and on the basis of the first model, removing the first full-connection layer of the first model, and adding a second full-connection layer according to the number of ship types of the slices with the ship types to construct the ship model.
In one embodiment of the invention, the first fully connected layer comprises a first Softmax layer and two first fully connected layers.
In another embodiment of the present invention, the second fully connected layer comprises a second Softmax layer and a second fully connected layer.
In a preferred embodiment of the invention, the first model is selected from VGG16, VGG19, inclusion v3 or Xception.
The above embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and the scope of the present invention is defined by the claims. Various modifications and equivalents may be made by those skilled in the art within the spirit and scope of the present invention, and such modifications and equivalents should also be considered as falling within the scope of the present invention.
Claims (6)
1. A method of training a convolutional neural network for SAR image ship classification, comprising:
acquiring a slice with a ship category in an SAR image;
training a convolutional neural network for SAR image ship classification based on the slice with the ship class so as to enable the convolutional neural network to reach preset training precision; the method specifically comprises the following steps:
setting initialization parameters of convolutional neural network training for SAR image ship classification;
inputting the slices with the ship categories into a convolutional neural network for SAR image ship classification to be trained to obtain an output result;
judging whether the output result reaches the preset training precision or not;
if not, adjusting initialization parameters, and returning to the step of inputting the slices with the ship categories into a convolutional neural network for SAR image ship classification to train so as to obtain an output result;
the output result reaches the preset training precision;
the convolution neural network for SAR image ship classification is constructed by removing a first full-connection layer of a first model on the basis of the first model and adding a second full-connection layer according to the number of ship types of the slices with the ship types;
setting initialization parameters for convolutional neural network training of SAR image ship classification, wherein the initialization parameters comprise pre-training parameters obtained by pre-training a first model and serving as the initialization parameters of the convolutional neural network for SAR image ship classification;
pre-training the first model on an optical data set to obtain pre-training parameters;
wherein the first model is selected from VGG16, VGG19, IncepotionV 3, or Xcenter; two types of methods are adopted to train the convolutional neural network for SAR image ship classification: the first training method is as follows: training all the connecting layers; the second type of training method is: only the second fully connected layer is trained.
2. The training method of claim 1, wherein the first fully-connected layer comprises a first Softmax layer and two first fully-connected layers, and the second fully-connected layer comprises a second Softmax layer and a second fully-connected layer.
3. The training method of claim 1, wherein the acquiring a slice with a ship class in the SAR image comprises:
and acquiring the section with the ship category according to ship information and a preset rule on the basis of the SAR image.
4. The training method of claim 1, wherein the training of the convolutional neural network for the ship classification of the SAR image based on the slice with the ship class comprises training all connection layers of the convolutional neural network for the ship classification of the SAR image, or
And training a second full connection layer of the convolutional neural network for SAR image ship classification.
5. A method for ship classification of SAR images based on the convolutional neural network obtained after training of the training method of the convolutional neural network for ship classification of SAR images in any one of claims 1 to 4, which comprises the following steps:
acquiring a ship slice in the SAR image to be classified;
and inputting the ship slice into a convolutional neural network which reaches the preset training precision and is used for SAR image ship classification to obtain a ship classification result of the ship slice.
6. A ship classification model is realized on the basis of a convolutional neural network which can reach preset training precision, wherein the convolutional neural network is constructed by removing a first full-connection layer of a first model on the basis of the first model and adding a second full-connection layer according to the number of ship classes of slices with the ship classes;
the method for enabling the convolutional neural network to achieve the preset training precision specifically comprises the following steps: setting initialization parameters of convolutional neural network training for SAR image ship classification; inputting the slices with the ship categories into a convolutional neural network for SAR image ship classification to be trained to obtain an output result; judging whether the output result reaches the preset training precision or not; if not, adjusting initialization parameters, and returning to the step of inputting the slices with the ship categories into a convolutional neural network for SAR image ship classification to train so as to obtain an output result; the output result reaches the preset training precision;
setting initialization parameters for convolutional neural network training of SAR image ship classification, wherein the initialization parameters comprise pre-training parameters obtained by pre-training a first model and serving as the initialization parameters of the convolutional neural network for SAR image ship classification;
pre-training the first model on an optical data set to obtain pre-training parameters;
wherein the first model is selected from VGG16, VGG19, IncepotionV 3, or Xcenter; two types of methods are adopted to train the convolutional neural network for SAR image ship classification: the first training method is as follows: training all the connecting layers; the second type of training method is: only the second fully connected layer is trained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810450109.3A CN108664933B (en) | 2018-05-11 | 2018-05-11 | Training method of convolutional neural network for SAR image ship classification, classification method of convolutional neural network and ship classification model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810450109.3A CN108664933B (en) | 2018-05-11 | 2018-05-11 | Training method of convolutional neural network for SAR image ship classification, classification method of convolutional neural network and ship classification model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108664933A CN108664933A (en) | 2018-10-16 |
| CN108664933B true CN108664933B (en) | 2021-12-28 |
Family
ID=63779348
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810450109.3A Active CN108664933B (en) | 2018-05-11 | 2018-05-11 | Training method of convolutional neural network for SAR image ship classification, classification method of convolutional neural network and ship classification model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108664933B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583412A (en) * | 2018-12-07 | 2019-04-05 | 中国科学院遥感与数字地球研究所 | A kind of training method and its ship detecting method carrying out ship detecting using convolutional neural networks |
| CN109766780A (en) * | 2018-12-20 | 2019-05-17 | 武汉理工大学 | A kind of ship smog emission on-line checking and method for tracing based on deep learning |
| CN110427981B (en) * | 2019-07-11 | 2023-01-31 | 四川大学 | SAR ship detection system and method based on deep neural network |
| CN110415224A (en) * | 2019-07-22 | 2019-11-05 | 北京金交信息通信导航设计院 | A kind of marine ships remote sense monitoring system and platform and method |
| CN110633353B (en) * | 2019-07-29 | 2020-05-19 | 南京莱斯网信技术研究院有限公司 | Ship type counterfeit monitoring method based on ensemble learning |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5057843A (en) * | 1990-06-25 | 1991-10-15 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Method for providing a polarization filter for processing synthetic aperture radar image data |
| CN106408030A (en) * | 2016-09-28 | 2017-02-15 | 武汉大学 | SAR image classification method based on middle lamella semantic attribute and convolution neural network |
| CN107563999A (en) * | 2017-09-05 | 2018-01-09 | 华中科技大学 | A kind of chip defect recognition methods based on convolutional neural networks |
| CN107563422A (en) * | 2017-08-23 | 2018-01-09 | 西安电子科技大学 | A kind of polarization SAR sorting technique based on semi-supervised convolutional neural networks |
| CN107886123A (en) * | 2017-11-08 | 2018-04-06 | 电子科技大学 | A kind of synthetic aperture radar target identification method based on auxiliary judgement renewal learning |
-
2018
- 2018-05-11 CN CN201810450109.3A patent/CN108664933B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5057843A (en) * | 1990-06-25 | 1991-10-15 | The United States Of America As Represented By The Administrator Of The National Aeronautics And Space Administration | Method for providing a polarization filter for processing synthetic aperture radar image data |
| CN106408030A (en) * | 2016-09-28 | 2017-02-15 | 武汉大学 | SAR image classification method based on middle lamella semantic attribute and convolution neural network |
| CN107563422A (en) * | 2017-08-23 | 2018-01-09 | 西安电子科技大学 | A kind of polarization SAR sorting technique based on semi-supervised convolutional neural networks |
| CN107563999A (en) * | 2017-09-05 | 2018-01-09 | 华中科技大学 | A kind of chip defect recognition methods based on convolutional neural networks |
| CN107886123A (en) * | 2017-11-08 | 2018-04-06 | 电子科技大学 | A kind of synthetic aperture radar target identification method based on auxiliary judgement renewal learning |
Non-Patent Citations (2)
| Title |
|---|
| 《Classification of very high resolution SAR image based on convolutional neural network》;Jinxin Li 等;《2017 International Workshop on Remote Sensing with Intelligent Processing (RSIP)》;20170626;全文 * |
| 《SAR图像船只分类识别研究进展》;吴樊 等;《遥感技术与应用》;20140215;第29卷(第1期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108664933A (en) | 2018-10-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108664933B (en) | Training method of convolutional neural network for SAR image ship classification, classification method of convolutional neural network and ship classification model | |
| CN108230329B (en) | Semantic segmentation method based on multi-scale convolution neural network | |
| Zhao et al. | Transfer learning with fully pretrained deep convolution networks for land-use classification | |
| Luus et al. | Multiview deep learning for land-use classification | |
| CN108399362A (en) | A kind of rapid pedestrian detection method and device | |
| CN110533631A (en) | SAR image change detection based on the twin network of pyramid pondization | |
| CN111681178B (en) | Knowledge distillation-based image defogging method | |
| CN109784148A (en) | Biopsy method and device | |
| CN110136162B (en) | Unmanned aerial vehicle visual angle remote sensing target tracking method and device | |
| CN104866868A (en) | Metal coin identification method based on deep neural network and apparatus thereof | |
| KR102440385B1 (en) | Method and apparatus of recognizing motion pattern base on combination of multi-model | |
| CN109919252A (en) | The method for generating classifier using a small number of mark images | |
| CN110245711A (en) | The SAR target identification method for generating network is rotated based on angle | |
| CN110097029B (en) | Identity authentication method based on high way network multi-view gait recognition | |
| CN109461177B (en) | Monocular image depth prediction method based on neural network | |
| CN113095333B (en) | Unsupervised feature point detection method and unsupervised feature point detection device | |
| CN106971189B (en) | A kind of noisy method for recognising star map of low resolution | |
| CN107016371A (en) | UAV Landing Geomorphological Classification method based on improved depth confidence network | |
| CN107766828A (en) | UAV Landing Geomorphological Classification method based on wavelet convolution neutral net | |
| CN111046915B (en) | Method for generating style characters | |
| CN113743417A (en) | Semantic segmentation method and semantic segmentation device | |
| CN116740344A (en) | Knowledge distillation-based lightweight remote sensing image semantic segmentation method and device | |
| CN112733777B (en) | Road extraction method, device and equipment of remote sensing image and storage medium | |
| CN113673369A (en) | Remote sensing image scene planning method and device, electronic equipment and storage medium | |
| CN106663186B (en) | method and system for face recognition |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20211124 Address after: 100190 No. 19 West North Fourth Ring Road, Haidian District, Beijing Applicant after: Research Institute of aerospace information innovation, Chinese Academy of Sciences Address before: No. 9 Dengzhuang South Road, Haidian District, Beijing 100094 Applicant before: Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |