CN114545354A - Sea surface target classification method and system - Google Patents

Sea surface target classification method and system Download PDF

Info

Publication number
CN114545354A
CN114545354A CN202210192424.7A CN202210192424A CN114545354A CN 114545354 A CN114545354 A CN 114545354A CN 202210192424 A CN202210192424 A CN 202210192424A CN 114545354 A CN114545354 A CN 114545354A
Authority
CN
China
Prior art keywords
recursive
sea surface
graph
detection data
training
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
Application number
CN202210192424.7A
Other languages
Chinese (zh)
Inventor
时艳玲
郭亚星
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Posts and Telecommunications
Original Assignee
Nanjing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing University of Posts and Telecommunications filed Critical Nanjing University of Posts and Telecommunications
Priority to CN202210192424.7A priority Critical patent/CN114545354A/en
Publication of CN114545354A publication Critical patent/CN114545354A/en
Pending legal-status Critical Current

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
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/417—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section involving the use of neural networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a sea surface target classification method and a system in the field of sea surface target classification, comprising the following steps: acquiring detection data of the sea surface through a radar, and deducing to obtain a detection recursion graph corresponding to the detection data; inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result; the training process of the classifier model comprises the following steps: acquiring historical detection data of the sea surface, and preprocessing the historical detection data to construct a recursion graph data set; training the convolutional neural network by using a recursive graph data set to obtain a classifier model with the classification accuracy rate larger than a set value; the invention generates a recursion map from the sea surface echo data, and utilizes the convolution neural network to perform autonomous feature extraction to distinguish the targets from the clutter, thereby realizing the rapid and accurate classification effect of the sea surface targets.

Description

Sea surface target classification method and system
Technical Field
The invention belongs to the field of sea surface target classification, and particularly relates to a sea surface target classification method and a sea surface target classification system.
Background
When the radar is used for sea surface target detection, the radar is often influenced by sea clutter. Because the radar cross-sectional area (RCS) of the small targets floating on the sea surface is small, and the echo energy is weak, accurate classification of the sea surface targets is always a difficult point. The traditional classification method based on statistics utilizes a statistical model to fit sea clutter data, but the statistical model cannot be well fitted due to the fact that sea clutter has the characteristics of nonlinearity, non-Gaussian, non-stability and the like.
The existing classification model extracts features by means of differences between clutter and target echoes, and then trains the classification model through machine learning means such as KNN, SVM and the like so as to achieve the purpose of classification. However, the extraction of the features has larger time complexity, and the existing classification model has subjectivity and limitation.
Disclosure of Invention
The invention aims to provide a sea surface target classification method and a system, which are used for generating a recursion graph from sea surface echo data, and performing autonomous feature extraction by using a convolutional neural network to distinguish a target from a clutter so as to realize a quick and accurate sea surface target classification effect.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
the invention provides a sea surface target classification method, which is characterized by comprising the following steps:
collecting detection data of the sea surface through a radar, and deducing to obtain a detection recursion graph corresponding to the detection data;
inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result;
the training process of the classifier model comprises the following steps:
acquiring historical detection data of the sea surface, and segmenting the historical detection data according to a set time length to obtain a detection data unit;
carrying out phase space reconstruction on each detection data unit to obtain an embedded vector, and constructing a recursive matrix according to the embedded vector; carrying out gray level normalization processing on the recursive matrix, converting the recursive matrix into a training recursive graph, and constructing a recursive graph data set;
and training the convolutional neural network by using a recursive graph data set to obtain a classifier model with the classification accuracy rate larger than a set value.
Preferably, the detection data is a sea echo amplitude sequence obtained by using radar.
Preferably, the method for performing phase space reconstruction on each detection data unit to obtain the embedded vector includes:
performing phase space reconstruction on the time sequence of the detection data unit to obtain an embedded vector after the phase space reconstruction, wherein an expression formula of the embedded vector is as follows:
v(r)＝{x(r),x(r+τ),x(r+2τ),...,x(r+(m-1)τ)}
where x (·) represents a time series, m represents an embedding dimension, τ represents an embedding delay, r represents a time sample, and v (r) represents an embedding vector.
Preferably, the method for constructing the recursive matrix according to the embedded vector comprises the following steps:
deriving a recursive matrix element based on the embedded vector, the expression formula being:
Figure BDA0003524858590000021
wherein r is more than or equal to 11≤n，1≤r2N, n representing the length of the time sample;
based on recursive matrix elements
Figure BDA0003524858590000022
Constructing a recursive matrix T, wherein the expression formula is as follows:
Figure BDA0003524858590000023
preferably, the method for performing the gray scale normalization process on the recursive matrix includes:
for recursive matrix elements
Figure BDA0003524858590000024
Carrying out gray level normalization treatment, wherein the gray level normalization formula is as follows:
Figure BDA0003524858590000031
wherein the content of the first and second substances,
Figure BDA0003524858590000032
expressed as transformed gray values, RminExpressed as the minimum grey value, R, of the recursive matrix element in the recursive matrix TmaxExpressed as the maximum grey value of the recursive matrix elements in the recursive matrix T;
in the recursive matrix TOf the recursive matrix element
Figure BDA0003524858590000033
Replacement by recursive matrix elements
Figure BDA0003524858590000034
A recursive matrix T' is constructed.
Preferably, the method for training the convolutional neural network by using the recursive graph data set comprises the following steps:
labeling targets and clutters in the training recursive graph through labels; dividing a recursion graph data set into a training set, a verification set and a test set according to a set proportion;
inputting the training set into a convolutional neural network for training, verifying the classification accuracy of the trained convolutional neural network through a verification set, and adjusting network parameters of the convolutional neural network;
inputting the test set into the verified convolutional neural network; when the classification accuracy rate is judged to be larger than a set value, outputting the convolutional neural network as a classifier model; and when the classification accuracy is judged to be less than or equal to the set value, retraining the convolutional neural network.
The invention provides a sea surface target classification system based on a convolutional neural network, which comprises the following components:
the acquisition module is used for acquiring detection data of the sea surface through a radar and deducing to obtain a detection recursion graph corresponding to the detection data;
the classification module is used for inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result;
the detection data unit acquisition module is used for acquiring historical detection data of the sea surface and segmenting the historical detection data according to a set time length to obtain a detection data unit;
the recursive graph data set construction module is used for carrying out phase space reconstruction on each detection data unit to obtain an embedded vector and constructing a recursive matrix according to the embedded vector; carrying out gray level normalization processing on the recursive matrix, converting the recursive matrix into a training recursive graph, and constructing a recursive graph data set;
and the training module is used for training the convolutional neural network by utilizing the recursive graph data set to obtain a classifier model with the classification accuracy rate larger than a set value.
Preferably, the classifier model includes convolutional layer C1, pooling layer S2, convolutional layer C3, pooling layer S4, convolutional layer C5, full-link layer F6, and output layer O7: the number of convolution kernels of the convolutional layer C1 is 16, and the number of convolution kernels of the convolutional layer C3 is 32; the convolutional layer C1, the convolutional layer C3 and the convolutional layer C5 are used for carrying out convolutional learning on the recursive graph; the pooling layers S2, S4 are used to take maximum pooling for the recursive graph; the recursive graph is processed by a convolutional layer C5 and then sequentially input into a full connection layer F6 and an output layer O7, and the output layer O7 outputs classification results.
A third aspect of the present invention provides a computer-readable storage medium, characterized in that a computer program is stored thereon, which, when being executed by a processor, carries out the steps of the method for classifying a sea surface object as set forth in any one of claims 1 to 7.
A fourth aspect of the present invention provides a computing device comprising one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods of sea surface object classification.
Compared with the prior art, the invention has the following beneficial effects:
the method comprises the steps of carrying out phase space reconstruction on each detection data unit to obtain an embedded vector, and constructing a recursive matrix according to the embedded vector; carrying out gray level normalization processing on the recursive matrix, converting the recursive matrix into a training recursive graph, and constructing a recursive graph data set; training the convolutional neural network through a recursive graph data set; the method autonomously extracts the features based on the recursion graph and the convolutional neural network, thereby avoiding the subjectivity and the limitation of feature extraction; meanwhile, the classifier model obtained based on the convolutional neural network training can have better generalization capability and robustness, and can obtain a good classification effect.
Drawings
Fig. 1 is a flowchart of a sea surface object classification method according to an embodiment of the present invention;
FIG. 2 is a flow chart of convolutional neural network training provided by an embodiment of the present invention;
fig. 3 is a histogram of evaluation indexes provided by the embodiment of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for illustrating the technical solutions of the present invention more clearly, and the protection scope of the present invention is not limited thereby.
Example one
As shown in fig. 1 to 2, a method for classifying sea surface objects includes:
acquiring detection data of the sea surface through a radar, wherein the detection data is a sea surface echo amplitude sequence obtained by the radar, the sea surface echo amplitude sequence is a time sequence formed by N pulse echo data received by the radar, and a detection recursion graph corresponding to the detection data is obtained through deduction;
inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result;
the training process of the classifier model comprises the following steps:
acquiring historical detection data of the sea surface, preprocessing the historical detection data, and constructing a recursion graph data set, wherein the process comprises the following steps:
segmenting historical detection data according to the time length of 1000 milliseconds to obtain detection data units; carrying out phase space reconstruction on each detection data unit to obtain an embedded vector; constructing a recursive matrix according to the embedded vector; and converting the recursive matrix into a training recursive graph after carrying out gray level normalization processing on the recursive matrix, and constructing a recursive graph data set.
The method for reconstructing the phase space of each detection data unit to obtain the embedded vector comprises the following steps:
performing phase space reconstruction on the time sequence of the detection data unit to obtain an embedded vector after the phase space reconstruction, wherein an expression formula of the embedded vector is as follows:
v(r)＝{x(r),x(r+τ),x(r+2τ),...,x(r+(m-1)τ)}
wherein x (·) represents a time series, m represents an embedding dimension, τ represents an embedding time delay, r represents a time sample, and v (r) represents an embedding vector; in this embodiment, the value of the embedding delay is 11, and the value of the embedding dimension is 5.
The method for constructing the recursive matrix according to the embedded vector comprises the following steps:
deriving a recursive matrix element based on the embedded vector, the expression formula being:
Figure BDA0003524858590000061
wherein r is more than or equal to 11≤n，1≤r2N, N ═ N- (m-1) τ, N representing the length of the time sample;
based on recursive matrix elements
Figure BDA0003524858590000062
Constructing a recursive matrix T, wherein the expression formula is as follows:
Figure BDA0003524858590000063
the method for converting the recursive matrix into the training recursive graph after carrying out gray level normalization processing comprises the following steps:
for recursive matrix elements
Figure BDA0003524858590000064
Carrying out gray level normalization treatment, wherein the gray level normalization formula is as follows:
Figure BDA0003524858590000065
wherein the content of the first and second substances,
Figure BDA0003524858590000066
expressed as transformed gray values, RminExpressed as the minimum grey value, R, of the recursive matrix element in the recursive matrix TmaxExpressed as the most recursive matrix elements in the recursive matrix TA large gray value;
combining recursive matrix elements in a recursive matrix T
Figure BDA0003524858590000067
Replacement by recursive matrix elements
Figure BDA0003524858590000068
Constructing a recursive matrix T'; and converting the recursion matrix T' into a training recursion graph, and constructing a recursion graph data set.
Labeling targets and clutters in the training recursive graph through labels; dividing a recursive graph data set into a training set, a verification set and a test set according to a ratio of 7:2: 1;
inputting the training set into a convolutional neural network for training, and transmitting the training set into a layer convolution layer of the convolutional neural network, wherein the calculation formula is as follows:
Figure BDA0003524858590000069
wherein f (-) is a non-linear activation function;
Figure BDA0003524858590000071
input for the first node of the convolutional layer; miThe range of local receptive field; k is a convolution kernel; bcIs the bias of the convolutional layer; the output of the layer convolution layer is a characteristic map of the recursion map.
Inputting the characteristic graph of the recursive graph into the next pooling layer, wherein the calculation formula is as follows:
yi＝fdown(χ1,χ2,...χL),χj∈P
wherein f isdown(. cndot.) is a down-sampling function, and the way of maximum pooling is used in the present invention. Chi shapejThe input of the ith node of the pooling layer is P, the local receptive field is P, and L is the number of nodes in the receptive field.
Inputting the output of the convolution layer and the pooling layer into the full-connection layer, wherein the calculation formula is as follows:
Figure BDA0003524858590000072
wherein o isfThe number of the characteristic picture elements;
Figure BDA0003524858590000073
for the input of the l-th node of the fully-connected layer, w and bfWeight matrix and bias of the full connection layer, respectively; typically the first fully-connected layer converts multiple features into vectors and the last fully-connected layer acts as the softmax classification layer.
Verifying the classification accuracy of the trained convolutional neural network through a verification set, and adjusting network parameters of the convolutional neural network;
inputting the test set into the verified convolutional neural network; when the classification accuracy rate is judged to be larger than a set value, outputting the convolutional neural network as a classifier model; and when the classification accuracy is judged to be less than or equal to the set value, retraining the convolutional neural network.
In the experiment, the method (RPs-CNN), Logistic Regression (LR), CART (Classification and Regression Trees, CART) and SVM are respectively used as comparison algorithms, the detection results of different algorithms under HH polarization are shown in fig. 3, and it can be seen from the figure that 3 indexes of accuracy, recall rate and F1 measurement of the method provided herein are all maintained at higher levels, wherein the most important F1 measurement can reach 92.05%, and the F1 measurements of LR, CART and SVM are respectively 79.00%, 66.19% and 73.50%. Therefore, the present invention is superior to other classification methods in classification performance.
Example two
A system for classifying sea surface targets based on convolutional neural network, which can apply the method for classifying sea surface targets according to the first embodiment, comprising:
the acquisition module is used for acquiring detection data of the sea surface through a radar and deducing to obtain a detection recursion graph corresponding to the detection data;
the classification module is used for inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result;
the detection data unit acquisition module is used for acquiring historical detection data of the sea surface and segmenting the historical detection data according to a set time length to obtain a detection data unit;
the recursive graph data set construction module is used for carrying out phase space reconstruction on each detection data unit to obtain an embedded vector and constructing a recursive matrix according to the embedded vector; carrying out gray level normalization processing on the recursive matrix, converting the recursive matrix into a training recursive graph, and constructing a recursive graph data set;
and the training module is used for training the convolutional neural network by utilizing the recursive graph data set to obtain a classifier model with the classification accuracy rate larger than a set value.
The classifier model includes convolutional layer C1, pooling layer S2, convolutional layer C3, pooling layer S4, convolutional layer C5, full-link layer F6, and output layer O7: the number of convolution kernels of the convolutional layer C1 is 16, and the number of convolution kernels of the convolutional layer C3 is 32; the convolution kernel size is 5 × 5. The pooling layer adopts a maximum pooling mode, the size of a pooling matrix is 2 multiplied by 2, and the step length is 2. The convolutional layer C1, the convolutional layer C3 and the convolutional layer C5 are used for carrying out convolutional learning on the recursive graph; the pooling layers S2, S4 are used to take maximum pooling for the recursive graph; the recursive graph is processed by a convolutional layer C5 and then sequentially input into a full connection layer F6 and an output layer O7, and the output layer O7 outputs classification results.
EXAMPLE III
A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method for classifying a sea surface object according to an embodiment.
Example four
A computing device comprising one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs comprising instructions for performing any of the sea surface object classification methods of embodiment one.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.

Claims (10)

1. A sea surface target classification method based on a convolutional neural network is characterized by comprising the following steps:
acquiring detection data of the sea surface through a radar, and deducing to obtain a detection recursion graph corresponding to the detection data;
inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result;
the training process of the classifier model comprises the following steps:
acquiring historical detection data of the sea surface, and segmenting the historical detection data according to a set time length to obtain a detection data unit;
carrying out phase space reconstruction on each detection data unit to obtain an embedded vector, and constructing a recursive matrix according to the embedded vector; carrying out gray level normalization processing on the recursive matrix, converting the recursive matrix into a training recursive graph, and constructing a recursive graph data set;
and training the convolutional neural network by using a recursive graph data set to obtain a classifier model with the classification accuracy rate larger than a set value.
2. The method of claim 1, wherein the detection data is a sequence of sea echo amplitudes obtained using radar.
3. The method of claim 1, wherein the step of performing a phase-space reconstruction on each detected data unit to obtain an embedded vector comprises:
performing phase space reconstruction on the time sequence of the detection data unit to obtain an embedded vector after the phase space reconstruction, wherein an expression formula of the embedded vector is as follows:
v(r)＝{x(r),x(r+τ),x(r+2τ),...,x(r+(m-1)τ)}
where x (·) represents a time series, m represents an embedding dimension, τ represents an embedding delay, r represents a time sample, and v (r) represents an embedding vector.
4. The method for classifying sea surface objects according to claim 3, wherein the method for constructing a recursive matrix according to the embedded vector comprises:
deriving a recursive matrix element based on the embedded vector, the expression formula being:
Figure FDA0003524858580000021
wherein r is more than or equal to 11≤n，1≤r2N, n representing the length of the time sample;
based on recursive matrix elements
Figure FDA0003524858580000022
Constructing a recursive matrix T, wherein the expression formula is as follows:
Figure FDA0003524858580000023
5. the method for classifying sea surface targets according to claim 4, wherein the method for performing gray-scale normalization processing on the recursive matrix comprises:
for recursive matrix elements
Figure FDA0003524858580000024
Carrying out gray level normalization treatment, wherein the gray level normalization formula is as follows:
Figure FDA0003524858580000025
wherein the content of the first and second substances,
Figure FDA0003524858580000026
expressed as transformed gray values, RminExpressed as the minimum grey value, R, of the recursive matrix element in the recursive matrix TmaxExpressed as the maximum grey value of the recursive matrix elements in the recursive matrix T;
combining recursive matrix elements in a recursive matrix T
Figure FDA0003524858580000027
Replacement by recursive matrix elements
Figure FDA0003524858580000028
A recursive matrix T' is constructed.
6. The method of classifying sea surface objects according to claim 1 or 5, wherein the method of training the convolutional neural network with the recursive graph data set comprises:
labeling targets and clutters in the training recursive graph through labels; dividing a recursion graph data set into a training set, a verification set and a test set according to a set proportion;
inputting the training set into a convolutional neural network for training, verifying the classification accuracy of the trained convolutional neural network through a verification set, and adjusting network parameters of the convolutional neural network;
inputting the test set into the verified convolutional neural network; when the classification accuracy rate is judged to be larger than a set value, outputting the convolutional neural network as a classifier model; and when the classification accuracy is judged to be less than or equal to the set value, retraining the convolutional neural network.
7. A convolutional neural network-based sea surface object classification system, comprising:
the acquisition module is used for acquiring detection data of the sea surface through a radar and deducing to obtain a detection recursion graph corresponding to the detection data;
the classification module is used for inputting the detection recursion graph into a classifier model trained in advance to obtain a classification result;
the detection data unit acquisition module is used for acquiring historical detection data of the sea surface and segmenting the historical detection data according to a set time length to obtain a detection data unit;
the recursive graph data set construction module is used for carrying out phase space reconstruction on each detection data unit to obtain an embedded vector and constructing a recursive matrix according to the embedded vector; carrying out gray level normalization processing on the recursive matrix, converting the recursive matrix into a training recursive graph, and constructing a recursive graph data set;
and the training module is used for training the convolutional neural network by utilizing the recursive graph data set to obtain a classifier model with the classification accuracy rate larger than a set value.
8. The method of classifying sea surface objects of claim 7, wherein the classifier model comprises a convolutional layer C1, a pooling layer S2, a convolutional layer C3, a pooling layer S4, a convolutional layer C5, a fully connected layer F6 and an output layer O7: the number of convolution kernels of the convolutional layer C1 is 16, and the number of convolution kernels of the convolutional layer C3 is 32; the convolutional layer C1, the convolutional layer C3 and the convolutional layer C5 are used for carrying out convolutional learning on the recursive graph; the pooling layers S2, S4 are used to take maximum pooling for the recursive graph; the recursive graph is processed by a convolutional layer C5 and then sequentially input into a full connection layer F6 and an output layer O7, and the output layer O7 outputs classification results.
9. Computer-readable storage medium, characterized in that it has stored thereon a computer program which, when being executed by a processor, carries out the steps of the method for classifying a sea surface object according to any one of claims 1 to 6.
10. A computing device, comprising,
one or more processors, memory, and one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods of claims 1-6.
CN202210192424.7A 2022-02-28 2022-02-28 Sea surface target classification method and system Pending CN114545354A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210192424.7A CN114545354A (en) 2022-02-28 2022-02-28 Sea surface target classification method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210192424.7A CN114545354A (en) 2022-02-28 2022-02-28 Sea surface target classification method and system

Publications (1)

Publication Number Publication Date
CN114545354A true CN114545354A (en) 2022-05-27

Family

ID=81662503

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210192424.7A Pending CN114545354A (en) 2022-02-28 2022-02-28 Sea surface target classification method and system

Country Status (1)

Country Link
CN (1) CN114545354A (en)

Similar Documents

Publication Publication Date Title
CN110472627B (en) End-to-end SAR image recognition method, device and storage medium
Kumar et al. Resnet-based approach for detection and classification of plant leaf diseases
CN109597043B (en) Radar signal identification method based on quantum particle swarm convolutional neural network
CN112001270B (en) Ground radar automatic target classification and identification method based on one-dimensional convolutional neural network
CN111126134B (en) Radar radiation source deep learning identification method based on non-fingerprint signal eliminator
Chen et al. Spatial–temporal convolutional gated recurrent unit network for significant wave height estimation from shipborne marine radar data
CN111080675A (en) Target tracking method based on space-time constraint correlation filtering
CN112395987A (en) SAR image target detection method based on unsupervised domain adaptive CNN
Nuhoglu et al. Image segmentation for radar signal deinterleaving using deep learning
Ye et al. Recognition algorithm of emitter signals based on PCA+ CNN
CN112116019A (en) Multi-sensor Vine Copula heterogeneous information decision fusion method
CN111832580A (en) SAR target identification method combining few-sample learning and target attribute features
Xu et al. Marine radar small target classification based on block-whitened time–frequency spectrogram and pre-trained CNN
Bahri et al. Time series forecasting using smoothing ensemble empirical mode decomposition and machine learning techniques
Ruan et al. Automatic recognition of radar signal types based on CNN-LSTM
CN105046707B (en) SAR image ship detection method based on n-order polynomial Function Fitting sea clutter
CN116953702A (en) Rotary target detection method and device based on deduction paradigm
CN114545354A (en) Sea surface target classification method and system
CN116030300A (en) Progressive domain self-adaptive recognition method for zero-sample SAR target recognition
Horn et al. Motion classification and height estimation of pedestrians using sparse radar data
Zhuangzhuang et al. Adaptive learning rate CNN for SAR ATR
CN108830290B (en) SAR image classification method based on sparse representation and Gaussian distribution
CN113421281A (en) Pedestrian micromotion part separation method based on segmentation theory
Lu et al. Image segmentation based on random neural network model and gabor filters
Nahi et al. A CNN-BiLSTM mixed model for moving targets’ classification from augmented micro-Doppler radar time-frequency distributions’ signatures

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