CN110232342A - Sea situation level determination method and device based on convolutional neural networks - Google Patents
Sea situation level determination method and device based on convolutional neural networks Download PDFInfo
- Publication number
- CN110232342A CN110232342A CN201910475458.5A CN201910475458A CN110232342A CN 110232342 A CN110232342 A CN 110232342A CN 201910475458 A CN201910475458 A CN 201910475458A CN 110232342 A CN110232342 A CN 110232342A
- Authority
- CN
- China
- Prior art keywords
- sea
- convolutional neural
- neural networks
- datagram
- data
- 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
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
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
-
- 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
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The embodiment of the present invention provides a kind of sea situation level determination method and device based on convolutional neural networks, the described method includes: obtaining the sea clutter datagram in sea area to be judged, the sea clutter datagram is built-up by the original sea clutter data in the sea area to be judged;By the sea clutter datagram in the sea area to be judged, it is input to default convolutional neural networks model, exports the sea situation grade in the sea area to be judged.Sea situation level determination method and device provided in an embodiment of the present invention based on convolutional neural networks, handles sea clutter datagram by trained convolutional neural networks model, obtains sea situation grade, and the efficiency and accuracy of the judgement of sea situation grade are improved.
Description
Technical field
The present invention relates to ocean monitoring technologytechnologies field more particularly to a kind of sea situation grades based on convolutional neural networks
Judgment method and device.
Background technique
Marine resources utilize to carry out etc. with offshore activities and are required to be monitored marine environment, wherein ocean surface is dynamic
Force parameter is a kind of important ocean environment parameter, mainly includes ocean near surface stream, wave and Ocean Wind-field etc..Unrestrained height is to use
The important parameter of extra large surface appearance is described, is the important content of wave forecast, common wave scale is defined.
In the prior art, for the measurement of sea situation grade, following two method is generallyd use:
One, scene point mensuration, mainly passes through the tools such as buoy, seat bottom type pressure sensor, subsurface buoy, current meter scene
Field survey.Two, remote sensing method connects sea clutter mechanism with oceanography, uses Bragg Resonance scattering mechanism reasonable dismissal
The single order interaction process of radio wave and wave establishes single order scattering section equation by further experiment and second order dissipates
Penetrate SECTION EQUATION.Ocean wave parameter is completely extracted from higher-frequency radar marine echo spectrum, it is necessary to inverting second order dispersion SECTION EQUATION.
The method of inverting ocean wave parameter can be divided into two classes: empirical formula method, directly be solved from marine echo spectrum using empirical equation
Unrestrained high and unrestrained period, such as Barrick method, Maresca method, Heron method;Spectral integral method, first using side linearly or nonlinearly
Method solves ocean wave spectrum from marine echo spectrum, then is integrated to obtain significant wave height and unrestrained cycle parameter to ocean wave spectrum, such as
Barrick method, Lipa method, Wyatt curve-fitting method etc..
But existing scene point mensuration, it can only obtain the data in specific time on local dotted line, to large area sea
The reflection of the case where domain is not comprehensive, is unable to get the marine environment data of real-time consecutive variations, measurement operation is by meteorological, sea situation item
The limitation of part, therefore, it is impossible to meet extensive, real time monitoring sea situation variation for a long time needs.And in existing remote sensing method, warp
Limitation of the unrestrained high measurement range of equation by working frequency is tested, integral equation method not can guarantee real-time, and to noise
It is higher than requiring, need more radars to measure elimination wave jointly to ambiguity.In addition, empirical formula method and integral equation method have respectively
Applicable sea situation range requires to carry out corresponding parameter adjustment, and further relate to according to radar system and marine environment difference
To the separation problem of first-order spectrum and second order spectrum, therefore, cause sea situation grade judging result accuracy low.
Summary of the invention
A kind of overcome the above problem the purpose of the embodiment of the present invention is that providing or at least be partially solved the above problem
Sea situation level determination method and device based on convolutional neural networks.
In order to solve the above-mentioned technical problem, on the one hand, the embodiment of the present invention provides a kind of sea based on convolutional neural networks
Condition level determination method, comprising:
The sea clutter datagram in sea area to be judged is obtained, the sea clutter datagram is by the original of the sea area to be judged
Sea clutter data are built-up;
By the sea clutter datagram in the sea area to be judged, be input to default convolutional neural networks model, output it is described to
Judge the sea situation grade in sea area.
Further, specific step is as follows for the acquisition default convolutional neural networks model:
Obtain multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample;
According to multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample, to mesh
Mark convolutional neural networks are trained, and obtain the default convolutional neural networks model.
Further, the sea clutter datagram for obtaining sea area to be judged, specifically includes:
The original sea clutter data in sea area to be judged are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, in the original sea clutter data
Ordinate of the column of element as pixel constructs described using the value of element in the original sea clutter data as pixel value
Sea clutter datagram.
Further, the original sea clutter data include the time-apart from 2-D data or azimuth-range 2-D data.
Further, described multiple sea clutter datagram samples of acquisition, specifically include:
The original sea clutter data in multiple sample sea areas are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, in the original sea clutter data
Ordinate of the column of element as pixel constructs multiple using the value of element in the original sea clutter data as pixel value
Original sea clutter datagram sample;
From each original sea clutter datagram sample, using preset window, several lesser figures are intercepted, as described
Sea clutter datagram sample.
Further, the target convolutional neural networks are LeNet convolutional neural networks.
Further, the size of the preset window is 10*10 to 300*300.
On the other hand, the embodiment of the present invention provides a kind of sea situation grade judgment means based on convolutional neural networks, comprising:
Obtain module, for obtaining the sea clutter datagram in sea area to be judged, the sea clutter datagram be by it is described to
Judge that the original sea clutter data in sea area are built-up;
Judgment module, for being input to default convolutional neural networks mould for the sea clutter datagram in the sea area to be judged
Type exports the sea situation grade in the sea area to be judged.
In another aspect, the embodiment of the present invention provides a kind of electronic equipment, comprising: memory, processor, and it is stored in institute
The computer program that can be run on memory and on the processor is stated, when the processor executes the computer program,
The step of realizing the above method.
Another aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating
Machine program, when the computer program is executed by processor, realize the above method the step of.
Sea situation level determination method and device provided in an embodiment of the present invention based on convolutional neural networks, by training
Convolutional neural networks model sea clutter datagram is handled, obtain sea situation grade, improve sea situation grade judgement effect
Rate and accuracy.
Detailed description of the invention
Fig. 1 is the sea situation level determination method schematic diagram provided in an embodiment of the present invention based on convolutional neural networks;
Fig. 2 is default convolutional neural networks model construction process schematic diagram provided in an embodiment of the present invention;
Fig. 3 is the process schematic of the data intercept sample provided in an embodiment of the present invention from data;
Fig. 4 is LeNet convolutional neural networks structural schematic diagram provided in an embodiment of the present invention;
Fig. 5 is the sea situation grade judgment means schematic diagram provided in an embodiment of the present invention based on convolutional neural networks;
Fig. 6 is the structural schematic diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to keep the purposes, technical schemes and advantages of the embodiment of the present invention clearer, implement below in conjunction with the present invention
Attached drawing in example, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment
It is a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiment of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Fig. 1 is the sea situation level determination method schematic diagram provided in an embodiment of the present invention based on convolutional neural networks, such as Fig. 1
Shown, the embodiment of the present invention provides a kind of sea situation level determination method based on convolutional neural networks, executing subject be based on
The sea situation grade judgment means of convolutional neural networks, this method comprises:
Step S101, the sea clutter datagram in sea area to be judged is obtained, the sea clutter datagram is by described wait judge
The original sea clutter data in sea area are built-up.
Specifically, by radar scanning sea area to be judged, the original of sea area to be judged is obtained from radar measured data
Sea clutter data.X-band radar can be used in the radar, for example, IPIX radar or CSIR's radar.
In the embodiment of the present invention, judgement is carried out to sea situation grade and uses pure sea clutter data, it is miscellaneous to get original sea
Wave number is pre-processed after, rejects land clutter, target echo and other interference echo data.
After the original sea clutter data for obtaining sea area to be judged, sea area to be judged is constructed using original sea clutter data
Sea clutter datagram.Original sea clutter data the characteristics of according to sea clutter data, are converted to sea clutter by the embodiment of the present invention
Datagram, to lay the foundation to carry out judgement to sea situation grade using convolutional neural networks model.
Step S102, by the sea clutter datagram in the sea area to be judged, it is input to default convolutional neural networks model, it is defeated
The sea situation grade in the sea area to be judged out.
Specifically, sea clutter has complex nonlinear and temporal correlation, and the embodiment of the present invention utilizes convolutional Neural net
Network has the characteristic of fitting effect well to the graph structure data of complex nonlinear feature, by the sea clutter number in sea area to be judged
According to figure, it is input to default convolutional neural networks model, exports the sea situation grade in sea area to be judged.The default convolutional neural networks mould
Type is to be trained using several clutter data sample sets to convolutional neural networks.Default convolutional neural networks model is real
The high-order for having showed the corresponding relationship between sea clutter data and sea situation grade indicates, presets the spy of convolutional neural networks model extraction
It is sufficiently high to levy Spatial Dimension, therefore, can be suitable under various radar systems and a variety of marine environment, and not by sea situation grade
Limitation.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks, passes through trained convolution
Neural network model handles sea clutter datagram, obtains sea situation grade, improves the efficiency and standard of the judgement of sea situation grade
True property.
Based on any of the above-described embodiment, further, the specific steps of the default convolutional neural networks model are obtained such as
Under:
Obtain multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample;
According to multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample, to mesh
Mark convolutional neural networks are trained, and obtain the default convolutional neural networks model.
Specifically, obtaining default convolutional neural networks model, specific step is as follows:
Firstly, obtaining multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample.
Mensuration is put by scene and obtains training sample set, which concentrates the sea clutter comprising multiple by number
Datagram sample, and to the corresponding sea situation grade that each sea clutter datagram sample is marked by label.
The quantity of training sample can be adjusted, for example, 1000 training samples of selection according to effect is realized.
Then, according to multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample,
Target convolutional neural networks are trained, default convolutional neural networks model is obtained.
That is, the accuracy rate using sample data is adjusted the network architecture parameters of target convolutional neural networks, finally
Obtain the default convolutional neural networks model judged for sea situation.These network architecture parameters include: each layer network structure
Learning rate, regularization term etc. when convolution kernel number, size and sliding step, pond mode, training.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks, passes through trained convolution
Neural network model handles sea clutter datagram, obtains sea situation grade, improves the efficiency and standard of the judgement of sea situation grade
True property.
Based on any of the above-described embodiment, further, the sea clutter datagram for obtaining sea area to be judged is specific to wrap
It includes:
The original sea clutter data in sea area to be judged are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, in the original sea clutter data
Ordinate of the column of element as pixel constructs described using the value of element in the original sea clutter data as pixel value
Sea clutter datagram.
Specifically, it after the original sea clutter data for obtaining sea area to be judged, is constructed using original sea clutter data
The sea clutter datagram in sea area to be judged.
Original sea clutter data are dimensional matrix data, using the row of element in original sea clutter data as the horizontal seat of pixel
Mark, using in original sea clutter data element column as pixel ordinate, using the value of element in original sea clutter data as
Pixel value can construct sea clutter datagram.
Original sea clutter data the characteristics of according to sea clutter data, are converted to sea clutter datagram by the embodiment of the present invention,
To lay the foundation to carry out judgement to sea situation grade using convolutional neural networks model.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks, passes through trained convolution
Neural network model handles sea clutter datagram, obtains sea situation grade, improves the efficiency and standard of the judgement of sea situation grade
True property.
Based on any of the above-described embodiment, further, the original sea clutter data include the time-apart from 2-D data or
Azimuth-range 2-D data.
Specifically, for radar under different working modes, the data of acquisition are different, and radar, which works in, stares mode
When, it is m- apart from 2-D data when the data of acquisition are;When radar works in scan pattern, the data of acquisition are azimuth-range two
Dimension data.
In actual use, fixed radar system, in the case that operating mode is constant, for pair in different sea surface observation regions
When answering the classification of sea situation grade, re -training model is not needed, it is only necessary to, will be wait sentence according to the judgment step in above-described embodiment
The sea clutter datagram in disconnected sea area, is input to default convolutional neural networks model, corresponding sea situation grade can be obtained.
In the operating mode or change classification task for changing radar, need to be constructed according to mode same with the above-mentioned embodiment
New convolutional neural networks model.
Original sea clutter data include the time-apart from 2-D data or azimuth-range 2-D data in the embodiment of the present invention.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks, can be to different radar moulds
Data under formula are handled, and application range is more wide.
Based on any of the above-described embodiment, further, described multiple sea clutter datagram samples of acquisition are specifically included:
The original sea clutter data in multiple sample sea areas are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, in the original sea clutter data
Ordinate of the column of element as pixel constructs multiple using the value of element in the original sea clutter data as pixel value
Original sea clutter datagram sample;
From each original sea clutter datagram sample, using preset window, several lesser figures are intercepted, as described
Sea clutter datagram sample.
Specifically, when obtaining multiple sea clutter datagram samples, sea clutter datagram size will affect sea
The accuracy of condition grade judgement.
Obtaining multiple sea clutter datagram samples, specific step is as follows:
Firstly, obtaining the original sea clutter data in multiple sample sea areas, original sea clutter data are dimensional matrix data, example
Such as, m- apart from 2-D data or azimuth-range 2-D data when.
Then, using the row of element in original sea clutter data as the abscissa of pixel, with first in original sea clutter data
Ordinate of the column of element as pixel constructs multiple original seas using the value of element in original sea clutter data as pixel value
Clutter data pattern sheet.
Finally, using preset window, intercepting several lesser figures from each original sea clutter datagram sample, make
For sea clutter datagram sample.
In order to make full use of the correlation of data, the square of preset window interception same size is slided according to preset step-length
Battle array saves the matrix intercepted every time as a sample, constitutes data set.If radar sea clutter data are in the time or apart from dimension ruler
It is very little larger, then slidably the preset window make interception sample between it is non-overlapping;If radar sea clutter data the time or away from
It is smaller from dimension size, then it can suitably reduce the step-length that window slides every time, can overlap between the sample of interception, to guarantee to cut
Take fully enough sample building training and test set.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks, by choosing suitable instruction
Practice sample, the result for judging sea situation grade is more accurate.
Based on any of the above-described embodiment, further, the target convolutional neural networks are LeNet convolutional neural networks.
Specifically, LeNet convolutional neural networks are made of multilayered nonlinear arithmetic element, by successively extracting effective spy
Sign, the high-order that structural information is obtained from a large amount of input datas indicates, can be used in that classification, recurrence, information retrieval etc. are specific to ask
Topic.
Use LeNet convolutional neural networks as the target for constructing default convolutional neural networks model in the embodiment of the present invention
Convolutional neural networks.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks uses LeNet convolution mind
Through network as the target convolutional neural networks for constructing default convolutional neural networks model, the result for judging sea situation grade is more
Accurately.
Based on any of the above-described embodiment, further, the size of the preset window is 10*10 to 300*300.
Specifically, when obtaining multiple sea clutter datagram samples, sea clutter datagram size will affect sea
The accuracy of condition grade judgement.
The size of preset window used in the embodiment of the present invention is 10*10 to 300*300.
Sea situation level determination method provided in an embodiment of the present invention based on convolutional neural networks is 10*10 using size
To the preset window of 300*300, to select training sample, the result for judging sea situation grade is more accurate.
Based on any of the above-described embodiment, Fig. 2 is default convolutional neural networks model construction mistake provided in an embodiment of the present invention
Journey schematic diagram, as shown in Figure 2.
The mentioned method of the present invention is needed using measured data building training set training network, by accurate on test set
Rate adjusts network architecture parameters, finally obtains the model that can be used for measured data sea situation grade separation.
In actual use, fixed radar system, in the case that operating mode is constant, for pair in different sea surface observation regions
When answering the classification of extra large state grade, re -training model is not needed, it is only necessary to 1 to step 3 construct data sample in the steps below, entirely
Portion obtains corresponding sea situation grade as test set input LeNet classification.Change operating mode or changes classification task
When, new data set training network need to be constructed in the same way.Several key steps will be described below.
Step 1: from (the radar work of the time obtained in radar measured data under sea situations at different levels/azimuth-range 2-D data
It is m- apart from 2-D data when being corresponded to when staring mode;Radar corresponds to azimuth-range 2-D data when working in scan pattern).
It include land clutter, target echo and other since neural metwork training and sea situation grade mark are directed to pure sea clutter data
The data of interference echo should be rejected as far as possible.For practical applications, then it can be obtained pure by the previously-scanned specified sea area of radar
Sea clutter data, for constructing data set.
Step 2: data intercept sample building training and test set from pure sea clutter data.Fig. 3 is the embodiment of the present invention
The process schematic of data intercept sample in the slave data provided, as shown in Figure 3.For high-resolution radar, sea clutter
When it is m- apparent texture structure can be shown on two dimensional gray figure, certain correlation is presented in the time and space.Cause
This, clutter data matrix is a kind of graph structure data, meets locality hypothesis and weight is shared it is assumed that having and utilizes convolutional Neural
The feasibility in theory that network is classified.In order to make full use of the correlation of data, in time and space dimension according to preset step-length
The matrix for sliding sliding window interception same size saves the matrix intercepted every time as a sample, constitutes data set.If Radar Sea
Clutter data when it is m- distance dimension size it is larger, then slidably sliding window make interception sample between it is non-overlapping;If Radar Sea is miscellaneous
Wave number according to the time or apart from dimension size it is smaller, then can suitably reduce the step-length that sliding window is slided every time, can have between the sample of interception
It partly overlaps, to guarantee to intercept enough sample building training and test set.
Step 3: according to significant wave height information corresponding with measured data, the data sample to intercept in step 2 determines mark
It signs and is numbered, stored respectively by training set with test set.The significant wave height model of each grade sea situation is determined by " Dow wave scale "
Significant wave height corresponding with measured data is enclosed, determines the affiliated sea situation classification of data intercept sample, and use tag identifier.The present invention is real
Applying the data set that example uses includes three sea situation grades: " SLIGHT " (3 grades), " MODERATE " (4 grades) and " ROUGH " (5 grades),
Correspond respectively to " 0.5-1.25m ", " 1.25-2.50m " and " 2.50-4.0m ".The data for randomly selecting wherein 20% are used as survey
Examination collection, remaining 80% is used as training set.
Step 4: the training set built input LeNet convolutional neural networks being trained, the standard on test set is passed through
True rate adjusts network architecture parameters, finally obtains the Artificial Neural Network Structures for sea situation classification.These parameters include, each
Learning rate, regularization term etc. when convolution kernel number, size and the sliding step of layer network structure, pond mode, training.This
Outside, some small network structure regulations are carried out for input and output layer also according to sample-size.
Fig. 4 is LeNet convolutional neural networks structural schematic diagram provided in an embodiment of the present invention, as shown in figure 4, other parameters
Setting, it should also according to the Output Size of each layer of network and loss function value decline situation be adjusted in the reasonable scope,
To reach convergence in minimum loss function value.After preservation model, input test collection is tested, and network query function is gone out
Sea situation grade classification result is compared with true sea situation grade, and statistical classification accuracy rate is simultaneously analyzed.
The embodiment of the present invention gives one kind and handles radar sea clutter measured data by LeNet convolutional neural networks, obtains
Into sea clutter data file when m- distance matrix corresponding to sea situation grade processing method, and pass through radar actual measurement sea
Clutter data is trained and tests, and demonstrates its feasibility and performance.In order to reach optimal training effect, for specific number
According to collection, suitable convolutional neural networks structural parameters need to be obtained by controlling the experiment of variable, formation adapts to the radar system
And operating mode, and there is the neural network model of separating capacity to a few class sea situation grades for including in input data.Experiment knot
Fruit shows that on IPIX data set, the differentiation accuracy rate to three/level Four sea situation is 95.754%, right on CSIR's modem
The differentiation accuracy rate of four/Pyatyi sea situation is 91.96%.This shows the method for the embodiment of the present invention to work in residing mode
It is lower that clutter sea situation grade separation is carried out with good feasibility using the polarized X-band radar of VV.Due to the two class numbers used
The clutter data of the different water areas of different moments acquisition is contained according to collection, it was demonstrated that this method has stronger generalization ability.
Based on any of the above-described embodiment, Fig. 5 is the sea situation grade provided in an embodiment of the present invention based on convolutional neural networks
Judgment means schematic diagram, as shown in figure 5, the embodiment of the present invention provides a kind of sea situation grade judgement dress based on convolutional neural networks
It sets, including obtains module 501 and judgment module 502, in which:
The sea clutter datagram that module 501 is used to obtain sea area to be judged is obtained, the sea clutter datagram is by described
The original sea clutter data in sea area to be judged are built-up;Judgment module 502 is used for the sea clutter number in the sea area to be judged
According to figure, it is input to default convolutional neural networks model, exports the sea situation grade in the sea area to be judged.
Specifically, by radar scanning sea area to be judged, the original of sea area to be judged is obtained from radar measured data
Sea clutter data.X-band radar can be used in the radar, for example, IPIX radar or CSIR's radar.
In the embodiment of the present invention, judgement is carried out to sea situation grade and uses pure sea clutter data, it is miscellaneous to get original sea
Wave number is pre-processed after, rejects land clutter, target echo and other interference echo data.
After the original sea clutter data for obtaining sea area to be judged, sea area to be judged is constructed using original sea clutter data
Sea clutter datagram.Original sea clutter data the characteristics of according to sea clutter data, are converted to sea clutter by the embodiment of the present invention
Datagram, to lay the foundation to carry out judgement to sea situation grade using convolutional neural networks model.
Sea clutter has complex nonlinear and temporal correlation, and the embodiment of the present invention is using convolutional neural networks to complicated non-
The graph structure data of linear character have the characteristic of fitting effect well, by the sea clutter datagram in sea area to be judged, input
To default convolutional neural networks model, the sea situation grade in sea area to be judged is exported.The default convolutional neural networks model is to utilize
What several clutter data sample sets were trained convolutional neural networks.It is extra large miscellaneous to preset convolutional neural networks model realization
Wave number is indicated according to the high-order of the corresponding relationship between sea situation grade, presets the feature space dimension of convolutional neural networks model extraction
It spends sufficiently high, therefore, can be suitable under various radar systems and a variety of marine environment, and not limited by sea situation grade.
Sea situation grade judgment means provided in an embodiment of the present invention based on convolutional neural networks, pass through trained convolution
Neural network model handles sea clutter datagram, obtains sea situation grade, improves the efficiency and standard of the judgement of sea situation grade
True property.
Based on any of the above-described embodiment, further, the sea situation grade judgment means of convolutional neural networks, further includes: instruction
Practice module.
The training module is corresponding for obtaining multiple sea clutter datagram samples and each sea clutter datagram sample
Sea situation grade;It is right according to multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample
Target convolutional neural networks are trained, and obtain the default convolutional neural networks model.
Based on any of the above-described embodiment, further, the acquisition module is specifically used for: obtaining the original in sea area to be judged
Beginning sea clutter data, the original sea clutter data are dimensional matrix data;With the row of element in the original sea clutter data
As the abscissa of pixel, using the column of element in the original sea clutter data as the ordinate of pixel, with the original sea
The value of element constructs the sea clutter datagram as pixel value in clutter data.
Based on any of the above-described embodiment, further, the original sea clutter data include the time-apart from 2-D data or
Azimuth-range 2-D data.
Based on any of the above-described embodiment, further, described multiple sea clutter datagram samples of acquisition are specifically included:
The original sea clutter data in multiple sample sea areas are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, in the original sea clutter data
Ordinate of the column of element as pixel constructs multiple using the value of element in the original sea clutter data as pixel value
Original sea clutter datagram sample;
From each original sea clutter datagram sample, using preset window, several lesser figures are intercepted, as described
Sea clutter datagram sample.
Based on any of the above-described embodiment, further, the target convolutional neural networks are LeNet convolutional neural networks.
Based on any of the above-described embodiment, further, the size of the preset window is 10*10 to 300*300.
Fig. 6 is the structural schematic diagram of electronic equipment provided in an embodiment of the present invention, as shown in fig. 6, the equipment includes: place
Device (processor) 601, memory (memory) 602, bus 603 are managed, and storage is on a memory and can be on a processor
The computer program of operation.
Wherein, processor 601 and memory 602 complete mutual communication by bus 603;
Processor 601 is for calling and executing the computer program in memory 602, to execute above-mentioned each method embodiment
In step, for example,
The sea clutter datagram in sea area to be judged is obtained, the sea clutter datagram is by the original of the sea area to be judged
Sea clutter data are built-up;
By the sea clutter datagram in the sea area to be judged, be input to default convolutional neural networks model, output it is described to
Judge the sea situation grade in sea area.
In addition, the logical order in above-mentioned memory can be realized and as independence by way of SFU software functional unit
Product when selling or using, can store in a computer readable storage medium.Based on this understanding, of the invention
Technical solution substantially the part of the part that contributes to existing technology or the technical solution can be with software in other words
The form of product embodies, which is stored in a storage medium, including some instructions use so that
One computer equipment (can be personal computer, server or the network equipment etc.) executes each embodiment institute of the present invention
State all or part of the steps of method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-
Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can be with
Store the medium of program code.
The embodiment of the present invention provides a kind of computer program product, and the computer program product is non-transient including being stored in
Computer program on computer readable storage medium, the computer program include program instruction, when described program instructs quilt
When computer executes, computer is able to carry out the step in above-mentioned each method embodiment, for example,
The sea clutter datagram in sea area to be judged is obtained, the sea clutter datagram is by the original of the sea area to be judged
Sea clutter data are built-up;
By the sea clutter datagram in the sea area to be judged, be input to default convolutional neural networks model, output it is described to
Judge the sea situation grade in sea area.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, is stored thereon with computer program, when
When the computer program is executed by processor, the step in above-mentioned each method embodiment is realized, for example,
The sea clutter datagram in sea area to be judged is obtained, the sea clutter datagram is by the original of the sea area to be judged
Sea clutter data are built-up;
By the sea clutter datagram in the sea area to be judged, be input to default convolutional neural networks model, output it is described to
Judge the sea situation grade in sea area.
The embodiments such as device and equipment described above are only schematical, wherein described be used as separate part description
Unit may or may not be physically separated, component shown as a unit may or may not be
Physical unit, it can it is in one place, or may be distributed over multiple network units.It can be according to the actual needs
Some or all of the modules therein is selected to achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying
In the case where creative labor, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features;
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of sea situation level determination method based on convolutional neural networks characterized by comprising
The sea clutter datagram in sea area to be judged is obtained, the sea clutter datagram is miscellaneous by the original sea in the sea area to be judged
Wave number is according to built-up;
By the sea clutter datagram in the sea area to be judged, it is input to default convolutional neural networks model, output is described wait judge
The sea situation grade in sea area.
2. the sea situation level determination method according to claim 1 based on convolutional neural networks, which is characterized in that obtain institute
Stating default convolutional neural networks model, specific step is as follows:
Obtain multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample;
According to multiple sea clutter datagram samples and the corresponding sea situation grade of each sea clutter datagram sample, to target volume
Product neural network is trained, and obtains the default convolutional neural networks model.
3. the sea situation level determination method according to claim 1 based on convolutional neural networks, which is characterized in that described to obtain
The sea clutter datagram for taking sea area to be judged, specifically includes:
The original sea clutter data in sea area to be judged are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, with element in the original sea clutter data
Ordinate of the column as pixel it is miscellaneous to construct the sea using the value of element in the original sea clutter data as pixel value
Wave datagram.
4. the sea situation level determination method according to claim 1 based on convolutional neural networks, which is characterized in that the original
Beginning sea clutter data include the time-apart from 2-D data or azimuth-range 2-D data.
5. the sea situation level determination method according to claim 2 based on convolutional neural networks, which is characterized in that described to obtain
Multiple sea clutter datagram samples are taken, are specifically included:
The original sea clutter data in multiple sample sea areas are obtained, the original sea clutter data are dimensional matrix data;
Using the row of element in the original sea clutter data as the abscissa of pixel, with element in the original sea clutter data
Ordinate of the column as pixel constructed multiple original using the value of element in the original sea clutter data as pixel value
Sea clutter datagram sample;
From each original sea clutter datagram sample, using preset window, several lesser figures are intercepted, it is miscellaneous as the sea
Wave data pattern sheet.
6. the sea situation level determination method according to claim 2 based on convolutional neural networks, which is characterized in that the mesh
Mark convolutional neural networks are LeNet convolutional neural networks.
7. the sea situation level determination method according to claim 5 based on convolutional neural networks, which is characterized in that described pre-
If the size of window is 10*10 to 300*300.
8. a kind of sea situation grade judgment means based on convolutional neural networks characterized by comprising
Module is obtained, for obtaining the sea clutter datagram in sea area to be judged, the sea clutter datagram is by described wait judge
The original sea clutter data in sea area are built-up;
Judgment module, it is defeated for being input to default convolutional neural networks model for the sea clutter datagram in the sea area to be judged
The sea situation grade in the sea area to be judged out.
9. a kind of electronic equipment, including memory, processor, and it is stored on the memory and can be on the processor
The computer program of operation, which is characterized in that when the processor executes the computer program, realize such as claim 1 to 7
The step of sea situation level determination method described in any one based on convolutional neural networks.
10. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that when the meter
When calculation machine program is executed by processor, realize that the sea situation grade as described in claim 1 to 7 is any based on convolutional neural networks is sentenced
The step of disconnected method.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910475458.5A CN110232342A (en) | 2019-06-03 | 2019-06-03 | Sea situation level determination method and device based on convolutional neural networks |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910475458.5A CN110232342A (en) | 2019-06-03 | 2019-06-03 | Sea situation level determination method and device based on convolutional neural networks |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110232342A true CN110232342A (en) | 2019-09-13 |
Family
ID=67859138
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910475458.5A Pending CN110232342A (en) | 2019-06-03 | 2019-06-03 | Sea situation level determination method and device based on convolutional neural networks |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110232342A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111738347A (en) * | 2020-06-28 | 2020-10-02 | 国家海洋环境预报中心 | Sea wave direction spectrum correction method and device, storage medium and electronic equipment |
CN111814960A (en) * | 2020-07-03 | 2020-10-23 | 国家气候中心 | Method and device for identifying weather climate system, storage medium and processor |
CN112068085A (en) * | 2020-10-16 | 2020-12-11 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Radar sea clutter original data rapid preprocessing method based on deep learning |
CN114519806A (en) * | 2022-01-29 | 2022-05-20 | 国家海洋环境预报中心 | Ocean wave level observation model training method and system |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109389058A (en) * | 2018-09-25 | 2019-02-26 | 中国人民解放军海军航空大学 | Sea clutter and noise signal classification method and system |
-
2019
- 2019-06-03 CN CN201910475458.5A patent/CN110232342A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109389058A (en) * | 2018-09-25 | 2019-02-26 | 中国人民解放军海军航空大学 | Sea clutter and noise signal classification method and system |
Non-Patent Citations (3)
Title |
---|
楚晓亮等: "利用X 波段雷达提取海浪信息的分析", 《中国海洋大学学报》 * |
赵巨波等: "海杂波统计特性分析", 《现代雷达》 * |
郝剑波: "基于深度学习的近岸海浪等级分类研究", 《中国优秀硕士学位论文全文数据库(电子期刊)基础科学辑》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111738347A (en) * | 2020-06-28 | 2020-10-02 | 国家海洋环境预报中心 | Sea wave direction spectrum correction method and device, storage medium and electronic equipment |
CN111738347B (en) * | 2020-06-28 | 2021-03-30 | 国家海洋环境预报中心 | Sea wave direction spectrum correction method and device, storage medium and electronic equipment |
CN111814960A (en) * | 2020-07-03 | 2020-10-23 | 国家气候中心 | Method and device for identifying weather climate system, storage medium and processor |
CN111814960B (en) * | 2020-07-03 | 2023-10-24 | 国家气候中心 | Weather and climate system identification method and device, storage medium and processor |
CN112068085A (en) * | 2020-10-16 | 2020-12-11 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Radar sea clutter original data rapid preprocessing method based on deep learning |
CN112068085B (en) * | 2020-10-16 | 2022-05-06 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Radar sea clutter original data rapid preprocessing method based on deep learning |
CN114519806A (en) * | 2022-01-29 | 2022-05-20 | 国家海洋环境预报中心 | Ocean wave level observation model training method and system |
CN114519806B (en) * | 2022-01-29 | 2022-10-11 | 国家海洋环境预报中心 | Ocean wave level observation model training method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110232342A (en) | Sea situation level determination method and device based on convolutional neural networks | |
CN109389058B (en) | Sea clutter and noise signal classification method and system | |
CN108052940A (en) | SAR remote sensing images waterborne target detection methods based on deep learning | |
CN108182450A (en) | A kind of airborne Ground Penetrating Radar target identification method based on depth convolutional network | |
CN110472627A (en) | One kind SAR image recognition methods end to end, device and storage medium | |
Buscombe et al. | Optical wave gauging using deep neural networks | |
CN109886217A (en) | A method of it is high that wave being detected from Nearshore Wave video based on convolutional neural networks | |
CN109683126A (en) | Direction of arrival measurement method, signal handling equipment and storage medium | |
CN109215034A (en) | A kind of Weakly supervised image, semantic dividing method for covering pond based on spatial pyramid | |
CN106228201A (en) | A kind of anti-Deceiving interference method of synthetic aperture radar based on shade characteristic | |
CN107256414A (en) | Polarimetric SAR Image convolutional neural networks sorting technique based on spatial feature collection of illustrative plates | |
Packalen et al. | Predicting the spatial pattern of trees by airborne laser scanning | |
CN108764006A (en) | A kind of SAR image object detection method based on deeply study | |
CN107607942B (en) | Based on the large scale electromagnetic scattering of deep learning model and the prediction technique of back scattering | |
CN107590515A (en) | The hyperspectral image classification method of self-encoding encoder based on entropy rate super-pixel segmentation | |
CN110728197B (en) | Single-tree-level tree species identification method based on deep learning | |
Wang et al. | Coastline interpretation from multispectral remote sensing images using an association rule algorithm | |
CN105549009A (en) | SAR image CFAR target detection method based on super pixels | |
Sundblad et al. | Comparing the ecological relevance of four wave exposure models | |
CN107527035A (en) | Earthquake damage to building information extracting method and device | |
CN110211127A (en) | Image partition method based on bicoherence network | |
CN111812598A (en) | Time domain and frequency domain multi-feature-based ground and sea clutter classification method | |
Zeng et al. | Wind speed inversion in high frequency radar based on neural network | |
CN110490155A (en) | A kind of no-fly airspace unmanned plane detection method | |
CN103425990A (en) | Synthetic aperture sonar target detection method based on textural characteristics |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190913 |
|
RJ01 | Rejection of invention patent application after publication |