CN114216464B - Intelligent underwater target positioning method - Google Patents
Intelligent underwater target positioning method Download PDFInfo
- Publication number
- CN114216464B CN114216464B CN202111351499.7A CN202111351499A CN114216464B CN 114216464 B CN114216464 B CN 114216464B CN 202111351499 A CN202111351499 A CN 202111351499A CN 114216464 B CN114216464 B CN 114216464B
- Authority
- CN
- China
- Prior art keywords
- underwater
- neural network
- underwater target
- data
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
-
- 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
- G01S5/00—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations
- G01S5/18—Position-fixing by co-ordinating two or more direction or position line determinations; Position-fixing by co-ordinating two or more distance determinations using ultrasonic, sonic, or infrasonic waves
-
- 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
Abstract
The invention relates to the technical field of underwater target positioning, and discloses an intelligent underwater target positioning method, which comprises the following steps: constructing an underwater target positioning mathematical model; generating a large amount of test simulation data by using the constructed underwater target positioning model, preprocessing the data, and taking the preprocessed data as training data; inputting training data as input to a neural network model; expanding a full-connection layer of the neural network, and training the parameters of the expanded full-connection layer by utilizing actual seabed positioning acquisition data based on a transfer learning algorithm to obtain a trained neural network model; and acquiring actual underwater data, taking the underwater data as input of a neural network model, and outputting a result as underwater target positioning coordinates. According to the invention, the training data is generated by establishing the underwater target positioning model, the neural network model is obtained by training through the transfer learning algorithm, and the problem of model environment mismatch is solved, so that the real-time positioning of the underwater target is realized.
Description
Technical Field
The invention relates to the technical field of underwater target positioning, in particular to an intelligent underwater target positioning method.
Background
As human exploration and development of the ocean becomes increasingly frequent, underwater acoustic positioning technology is variously developed in the fields of submarine optical cable line laying, large-scale ocean space and environment data acquisition, ocean resource exploration and development and the like. The global positioning system realizes the positioning of land equipment by using electromagnetic waves, and the GPS can not be directly applied to the water because the seawater can absorb the electromagnetic waves well, so that the underwater target positioning becomes a hot topic of the current research.
The signal is subject to the influence of complex marine environment, attenuation and distortion are easy to generate in the underwater propagation process by taking sound waves as carriers, refraction or reflection phenomena occur when the signal encounters obstacles, the original waveform of the signal is also influenced, and difficulty is brought to the underwater target positioning detection based on underwater acoustic signals.
In view of this, how to realize underwater target positioning by using acoustic signals is a problem to be solved by those skilled in the art.
Disclosure of Invention
The invention provides an intelligent underwater target positioning method, which aims at (1) optimizing an underwater sound wave transmission method; and (2) realizing real-time positioning of the underwater target.
The invention provides an intelligent underwater target positioning method, which comprises the following steps:
s1: constructing an underwater target positioning mathematical model;
s2: generating a large amount of test simulation data by using the constructed underwater target positioning model, preprocessing the data, and taking the preprocessed data as training data;
s3: inputting training data as input into a neural network model, and training the neural network model by utilizing an ADAM algorithm to obtain trained convolution-pooling layer parameters;
s4: fixing convolution-pooling layer parameters of the neural network model, expanding a full-connection layer of the neural network, and training the expanded full-connection layer parameters by utilizing actual seabed positioning acquisition data based on a migration learning algorithm to obtain a trained neural network model;
s5: the method comprises the steps of collecting actual underwater data, preprocessing the collected underwater data, taking the preprocessed underwater data as input of a neural network model, and outputting an output result as underwater target positioning coordinates.
As a further improvement of the present invention:
and in the step S1, a signal receiving model in an underwater target positioning mathematical model is constructed, and the method comprises the following steps:
setting M signal transmitting array elements and N signal receiving array elements on water, wherein the signal transmitting array elements transmit sound rays to an underwater target, and a signal receiving model of the signal receiving array elements is as follows:
wherein:
S ij (t) when the ith signal transmitting array element transmits signals, the jth signal receiving array element receives signals, and t represents signal transmitting time;
c i (t) represents a signal transmitted by an ith signal transmitting element;
representing a convolution operation;
n (t) represents an underwater noise signal;
h ij (t) represents an impulse response function of the underwater signal channel;
(x, y) represents the two-dimensional coordinates of the underwater target, the underwater target receives L incident sound rays, the signal receiving array element receives Q scattered sound rays, and the impulse response of the incident sound rays p to the underwater target is G p (x, y) with a delay of T p P is more than or equal to 1 and less than or equal to L, and the impulse response of scattered sound rays reaching signal receiving array elements is G q (x, y) with a delay of T q ，1≤q≤Q；
Representing impulse response formed by L incident sound rays and Q scattered sound rays according to time delay T in time domain p +T q Superimposed together to form the impulse response function h of the underwater target i ′ j (t)。
In the step S1, an underwater target positioning mathematical model is constructed according to the constructed signal receiving model, and the method comprises the following steps:
pulse compression processing is carried out on the received signal:
wherein:
* Representing complex conjugate;
D ij (t) represents the received signal S ij A pulse compression waveform result of (t);
constructing a mathematical model of a pulse compression waveform result and underwater target positioning coordinates:
wherein:
(x, y) represents the two-dimensional coordinates of the underwater target.
And in the step S2, generating a large amount of test simulation data by using the constructed underwater target positioning mathematical model, wherein the method comprises the following steps of:
according to the constructed underwater target positioning mathematical model, simulating signal transmitting array elements, signal receiving array elements and an underwater environment, randomly generating a plurality of underwater targets, transmitting signals to the underwater targets by the simulating signal transmitting array elements, calculating an underwater impulse response function, taking the underwater impulse response function, the underwater targets and the transmitting signals as parameters of the underwater positioning mathematical model, and obtaining an underwater target d k Corresponding pulse compression waveform result D k (t) where k.epsilon.1, E]E represents the number of generated underwater targets, d k Is positioned at the coordinate I k ＝(x k ,y k ) To be underwater target d k Position information I of (a) k Corresponding pulse compression waveform results D k (t) as a set of test simulation data, the test simulation data set is { (I) 1 ,D 1 (t)),(I 2 ,D 2 (t)),(I 3 ,D 3 (t)),…,(I E ,D E (t))}。
And S2, preprocessing the generated test simulation data, taking the preprocessed test simulation data as training data, wherein the method comprises the following steps of:
pre-emphasis pretreatment is carried out on pulse compression waveform results in a test simulation data set, wherein the pre-emphasis pretreatment formula is as follows:
D′ i (t)＝D i (t)-aD i (t-1)
wherein:
a represents a pre-emphasis coefficient, which is set to 0.98;
D i (t) represents an underwater target d i The corresponding pulse compression waveform results;
D′ i (t) is the pulse compression waveform result after pre-emphasis treatment;
the test simulation data set after pretreatment { (I) 1 ,D′ 1 (t)),(I 2 ,D′ 2 (t)),…,(I E ,D′ E (t)) as a training data set.
And S3, taking training data as model input to construct a neural network model, wherein the step comprises the following steps:
inputting pulse compression waveform results in training data as a neural network model, outputting underwater target position information in the training data as a neural network model, constructing a neural network model, wherein the constructed neural network model consists of two convolution-pooling layers and two full-connection layers, and the loss function of the neural network model is as follows:
wherein:
e represents the number of underwater targets in the training data;
I e representing an underwater target d e Is a real position information of the mobile terminal;
underwater target d representing neural network model output e Is a part of the position information of the mobile terminal;
in the step S3, training a neural network model by using an ADAM algorithm to obtain trained convolution-pooling layer parameters, wherein the training step comprises the following steps:
training a convolution-pooling layer in a neural network model by using an ADAM algorithm to obtain trained parameters of the convolution-pooling layer, wherein the ADAM algorithm comprises the following steps:
1) Initializing convolution-pooling layer parameters θ 0 Partial first moment estimate m 0 =0, partial moment estimate v 0 =0, h is 0;
2)h＝h+1；
3)wherein g h Representing the neural network model parameter as theta h Is a loss function gradient of f (θ) h -1) represents a neural network model parameter θ h A neural network model f θ Representing a Softmax activation function;
4)m h ＝β 1 m h-1 +(1-β 1 )g h wherein beta is 1 An exponential decay rate for moment estimation, set to 0.82;
5)v h ＝β 2 v h-1 +(1-β 2 )*(g h ) 2 wherein beta is 2 An exponential decay rate for the moment estimation, set to 0.79;
6) The update formula of the convolution-pooling layer parameters is:
wherein:
alpha is the learning rate, and is set to 0.6;
7) Repeating the steps 2) -6) until the parameters of the convolution-pooling layer are converged, and performing convolution-pooling after convergenceLayer parameter θ * 。
And S4, fixing convolutional-pooling layer parameters of the neural network model, expanding a fully-connected layer of the neural network, training the expanded fully-connected layer parameters by utilizing actual seabed positioning acquisition data based on transfer learning, and comprising the following steps:
taking the trained convolution-pooling layer parameters as the convolution-pooling layer parameters of the neural network model, and fixing the convolution-pooling layer parameters of the neural network model;
the method comprises the steps of acquiring submarine positioning acquisition data, wherein the submarine positioning acquisition data are submarine target positioning information and acoustic wave information, expanding a full connection layer into three layers, and the full connection layer is as follows:
I＝f(w k D′+b k )k＝1,2,3
wherein:
d' is input acoustic wave information, namely a pulse compression waveform result;
w k the weight of the k-th full-connection layer;
b k the offset of the k-th full-connection layer;
f (·) is the Relu activation function;
i represents the underwater target position information output by the neural network;
based on a training data set, training by utilizing an ADAM algorithm to obtain parameters of three full-connection layers, and fixing full-connection parameters of a first layer and a second layer, wherein the full-connection layer parameters comprise weights and offset of the full-connection layers;
the acoustic wave information in the submarine location acquisition data is used as the input of a neural network model, and the parameters of a third full-connection layer are adjusted by comparing the submarine target location information output by the neural network model with the submarine target location information, so that the distance between the two location information is keptMinimum, where I h Indicating seabed target positioning information in seabed positioning acquisition data, < +.>And the underwater target position information output by the neural network model is represented.
In the step S5, the underwater data of the actual underwater target positioning is collected, the collected underwater data is preprocessed, the preprocessed underwater data is used as the input of a neural network model, and the positioning coordinates of the underwater target are obtained, and the method comprises the following steps:
the method comprises the steps of collecting underwater data of actual underwater target positioning, preprocessing the collected underwater data to obtain a pulse compression waveform result of the underwater data, taking the pulse compression waveform result as input of a neural network model, and outputting underwater target positioning coordinates by using the neural network model.
Compared with the prior art, the invention provides an intelligent underwater target positioning method, which has the following advantages:
firstly, the scheme provides an underwater signal receiving model, which converts an underwater sound wave into a mathematical model, so that the problem of inaccurate positioning caused by attenuation and distortion of the sound wave in the underwater propagation process is avoided, wherein the signal receiving model of a signal receiving array element is as follows:
wherein: s is S ij (t) when the ith signal transmitting array element transmits signals, the jth signal receiving array element receives signals, and t represents signal transmitting time; c i (t) represents a signal transmitted by an ith signal transmitting element;representing a convolution operation; n (t) represents an underwater noise signal; h is a ij (t) represents an impulse response function of the underwater signal channel; (x, y) represents the two-dimensional coordinates of the underwater target, L incident sound rays are received by the underwater target, Q is received by the signal receiving array elementThe impulse response of the incident sound ray p to the underwater target is G p (x, y) with a delay of T p P is more than or equal to 1 and less than or equal to L, and the impulse response of scattered sound rays reaching signal receiving array elements is G q (x, y) with a delay of T q ，1≤q≤Q；/>Representing impulse response formed by L incident sound rays and Q scattered sound rays according to time delay T in time domain p +T q Overlapping to form impulse response function of underwater target; pulse compression processing is carried out on the received signals to obtain pulse compression waveforms of underwater sound waves, so that a mathematical model of a pulse compression waveform result and underwater target positioning coordinates is constructed:
wherein: (x, y) represents the two-dimensional coordinates of the underwater target. According to the constructed underwater target positioning mathematical model, simulating signal transmitting array elements, signal receiving array elements and an underwater environment, randomly generating a plurality of underwater targets, transmitting signals to the underwater targets by the simulating signal transmitting array elements, calculating an underwater impulse response function, taking the underwater impulse response function, the underwater targets and the transmitting signals as parameters of the underwater positioning mathematical model, and obtaining an underwater target d k Corresponding pulse compression waveform result D k (t) where k.epsilon.1, E]E represents the number of generated underwater targets, d k Is positioned at the coordinate I k ＝(x k ,y k ) To be underwater target d k Position information I of (a) k Corresponding pulse compression waveform results D k (t) as a set of test simulation data, the test simulation data set is { (I) 1 ,D 1 (t)),(I 2 ,D 2 (t)),(I 3 ,D 3 (t)),…,(I E ,D E (t)) thereby obtaining a large amount of training data, and training to obtain a more optimal neural network model for underwater target positioning.
Meanwhile, the scheme trains and obtains the full-connection layer of the neural network model by utilizing the migration learning method, and expands the full-connection layer into three layers by acquiring the seabed positioning acquisition data, wherein the full-connection layer is as follows:
I＝f(w k D′+b k )k＝1,2,3
wherein: d' is input acoustic wave information, namely a pulse compression waveform result; w (w) k The weight of the k-th full-connection layer; b k The offset of the k-th full-connection layer; f (·) is the Relu activation function; i represents the underwater target position information output by the neural network; based on a training data set, training by utilizing an ADAM algorithm to obtain parameters of three full-connection layers, and fixing full-connection parameters of a first layer and a second layer, wherein the full-connection layer parameters comprise weights and offset of the full-connection layers; the acoustic wave information in the submarine location acquisition data is used as the input of a neural network model, and the parameters of a third full-connection layer are adjusted by comparing the submarine target location information output by the neural network model with the submarine target location information, so that the distance between the two location information is keptMinimum, where I h Indicating seabed target positioning information in seabed positioning acquisition data, < +.>The method and the device solve the problem of model environment mismatch, namely the problem that a model obtained by training by using training data is possibly inapplicable in an actual scene by utilizing actual data to adjust parameters of a full-connection layer.
Drawings
FIG. 1 is a schematic flow chart of an intelligent underwater target positioning method according to an embodiment of the present invention;
the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
S1: and constructing an underwater target positioning mathematical model.
And in the step S1, a signal receiving model in an underwater target positioning mathematical model is constructed, and the method comprises the following steps:
setting M signal transmitting array elements and N signal receiving array elements on water, wherein the signal transmitting array elements transmit sound rays to an underwater target, and a signal receiving model of the signal receiving array elements is as follows:
wherein:
S ij (t) when the ith signal transmitting array element transmits signals, the jth signal receiving array element receives signals, and t represents signal transmitting time;
c i (t) represents a signal transmitted by an ith signal transmitting element;
representing a convolution operation;
n (t) represents an underwater noise signal;
h ij (t) represents an impulse response function of the underwater signal channel;
(x, y) represents the two-dimensional coordinates of the underwater target, the underwater target receives L incident sound rays, the signal receiving array element receives Q scattered sound rays, and the impulse response of the incident sound rays p to the underwater target is G p (x, y) with a delay of T p P is more than or equal to 1 and less than or equal to L, and the impulse response of scattered sound rays reaching signal receiving array elements is G q (x, y) with a delay of T q ，1≤q≤Q；
Representing impulse response formed by L incident sound rays and Q scattered sound rays according to time delay T in time domain p +T q Superimposed together to form the impulse response function h 'of the underwater target' ij (t)。
In the step S1, an underwater target positioning mathematical model is constructed according to the constructed signal receiving model, and the method comprises the following steps:
pulse compression processing is carried out on the received signal:
wherein:
* Representing complex conjugate;
D ij (t) represents the received signal S ij A pulse compression waveform result of (t);
constructing a mathematical model of a pulse compression waveform result and underwater target positioning coordinates:
wherein:
(x, y) represents the two-dimensional coordinates of the underwater target.
S2: generating a large amount of test simulation data by using the constructed underwater target positioning model, preprocessing the data, and taking the preprocessed data as training data.
And in the step S2, generating a large amount of test simulation data by using the constructed underwater target positioning mathematical model, wherein the method comprises the following steps of:
according to the constructed underwater target positioning mathematical model, simulating signal transmitting array elements, signal receiving array elements and an underwater environment, randomly generating a plurality of underwater targets, transmitting signals to the underwater targets by the simulating signal transmitting array elements, calculating an underwater impulse response function, taking the underwater impulse response function, the underwater targets and the transmitting signals as parameters of the underwater positioning mathematical model, and obtaining an underwater target d k Corresponding pulse compression waveform result D k (t) where k.epsilon.1, E]E represents the number of generated underwater targets, d k Is positioned at the coordinate I k ＝(x k ，y k ) To be underwater target d k Position information I of (a) k Corresponding pulse compression waveform results D k (t) as a set of test simulation data, the test simulation data set is { (I) 1 ，D 1 (t))，(I 2 ，D 2 (t))，(I 3 ，D 3 (t))，…，(I E ，D E (t))}。
And S2, preprocessing the generated test simulation data, taking the preprocessed test simulation data as training data, wherein the method comprises the following steps of:
pre-emphasis pretreatment is carried out on pulse compression waveform results in a test simulation data set, wherein the pre-emphasis pretreatment formula is as follows:
D′ i (t)＝D i (t)-aD i (t-1)
wherein:
a represents a pre-emphasis coefficient, which is set to 0.98;
D i (t) represents an underwater target d i The corresponding pulse compression waveform results;
D′ i (t) is the pulse compression waveform result after pre-emphasis treatment;
the test simulation data set after pretreatment { (I) 1 ,D′ 1 (t)),(I 2 ,D′ 2 (t)),…,(I E ,D′ E (t)) as a training data set.
S3: and taking the training data as input, inputting the training data into a neural network model, and training the neural network model by utilizing an ADAM algorithm to obtain the trained convolution-pooling layer parameters.
And S3, taking training data as model input to construct a neural network model, wherein the step comprises the following steps:
inputting pulse compression waveform results in training data as a neural network model, outputting underwater target position information in the training data as a neural network model, constructing a neural network model, wherein the constructed neural network model consists of two convolution-pooling layers and two full-connection layers, and the loss function of the neural network model is as follows:
wherein:
e represents the number of underwater targets in the training data;
I e representing an underwater target d e Is a real position information of the mobile terminal;
underwater target d representing neural network model output e Is a part of the position information of the mobile terminal;
in the step S3, training a neural network model by using an ADAM algorithm to obtain trained convolution-pooling layer parameters, wherein the training step comprises the following steps:
training a convolution-pooling layer in a neural network model by using an ADAM algorithm to obtain trained parameters of the convolution-pooling layer, wherein the ADAM algorithm comprises the following steps:
1) Initializing convolution-pooling layer parameters θ 0 Partial first moment estimate m 0 =0, partial moment estimate v 0 =0, h is 0;
2)h＝h+1；
3)wherein g h Representing the neural network model parameter as theta h Is a loss function gradient of f (θ) h -1) represents a neural network model parameter θ h A neural network model f θ Representing a Softmax activation function;
4)m h ＝β 1 m h-1 +(1-β 1 )g h wherein beta is 1 An exponential decay rate for moment estimation, set to 0.82;
5)v h ＝β 2 v h-1 +(1-β 2 )*(g h ) 2 wherein beta is 2 An exponential decay rate for the moment estimation, set to 0.79;
6) The update formula of the convolution-pooling layer parameters is:
wherein:
alpha is the learning rate, and is set to 0.6;
7) Repeating the steps 2) -6) until the parameters of the convolution-pooling layer are converged, and enabling the parameters of the converged convolution-pooling layer to be theta.
S4: and fixing convolution-pooling layer parameters of the neural network model, expanding a full-connection layer of the neural network, and training the expanded full-connection layer parameters by utilizing actual seabed positioning acquisition data based on a migration learning algorithm to obtain a trained neural network model.
And S4, fixing convolutional-pooling layer parameters of the neural network model, expanding a fully-connected layer of the neural network, training the expanded fully-connected layer parameters by utilizing actual seabed positioning acquisition data based on transfer learning, and comprising the following steps:
taking the trained convolution-pooling layer parameters as the convolution-pooling layer parameters of the neural network model, and fixing the convolution-pooling layer parameters of the neural network model;
the method comprises the steps of acquiring submarine positioning acquisition data, wherein the submarine positioning acquisition data are submarine target positioning information and acoustic wave information, expanding a full connection layer into three layers, and the full connection layer is as follows:
I＝f(w k D′+b k )k＝1,2,3
wherein:
d' is input acoustic wave information, namely a pulse compression waveform result;
w k the weight of the k-th full-connection layer;
b k the offset of the k-th full-connection layer;
f (·) is the Relu activation function;
i represents the underwater target position information output by the neural network;
based on a training data set, training by utilizing an ADAM algorithm to obtain parameters of three full-connection layers, and fixing full-connection parameters of a first layer and a second layer, wherein the full-connection layer parameters comprise weights and offset of the full-connection layers;
the acoustic wave information in the submarine location acquisition data is used as the input of a neural network model, and the parameters of a third full-connection layer are adjusted by comparing the submarine target location information output by the neural network model with the submarine target location information, so that the distance between the two location information is keptMinimum, where I h Indicating seabed target positioning information in seabed positioning acquisition data, < +.>And the underwater target position information output by the neural network model is represented.
S5: the method comprises the steps of collecting actual underwater data, preprocessing the collected underwater data, taking the preprocessed underwater data as input of a neural network model, and outputting an output result as underwater target positioning coordinates.
In the step S5, the underwater data of the actual underwater target positioning is collected, the collected underwater data is preprocessed, the preprocessed underwater data is used as the input of a neural network model, and the positioning coordinates of the underwater target are obtained, and the method comprises the following steps:
the method comprises the steps of collecting underwater data of actual underwater target positioning, preprocessing the collected underwater data to obtain a pulse compression waveform result of the underwater data, taking the pulse compression waveform result as input of a neural network model, and outputting underwater target positioning coordinates by using the neural network model.
It should be noted that, the foregoing reference numerals of the embodiments of the present invention are merely for describing the embodiments, and do not represent the advantages and disadvantages of the embodiments. And the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, apparatus, article or method that comprises the element.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) as described above, comprising instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method according to the embodiments of the present invention.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (7)
1. An intelligent underwater target positioning method, which is characterized by comprising the following steps:
s1: constructing an underwater target positioning mathematical model;
s2: generating a large amount of test simulation data by using the constructed underwater target positioning mathematical model, preprocessing the generated test simulation data, and taking the preprocessed test simulation data as training data;
s3: inputting training data as input into a neural network model, and training the neural network model by utilizing an ADAM algorithm to obtain trained convolution-pooling layer parameters;
the flow of ADAM algorithm is:
1) Initializing convolution-pooling layer parameters θ 0 Partial first moment estimate m 0 =0, partial moment estimate v 0 =0, h is 0;
2)h＝h+1；
3)wherein g h Representing the neural network model parameter as theta h Is a loss function gradient of f (θ) h-1 ) Representing the neural network model parameter as theta h A neural network model f θ Representing a Softmax activation function;
4)m h ＝β 1 m h-1 +(1-β 1 )g h wherein beta is 1 An exponential decay rate for moment estimation, set to 0.82;
5)v h ＝β 2 v h-1 +(1-β 2 )*(g h ) 2 wherein beta is 2 An exponential decay rate for the moment estimation, set to 0.79;
6) The update formula of the convolution-pooling layer parameters is:
wherein:
alpha is the learning rate, and is set to 0.6;
7) Repeating the steps 2) -6) until the parameters of the convolution-pooling layer are converged, and the parameters of the converged convolution-pooling layer are theta * ；
S4: fixing convolution-pooling layer parameters of a neural network model, expanding a full-connection layer of the neural network, training the expanded full-connection layer parameters by utilizing actual seabed positioning acquisition data based on a migration learning algorithm to obtain a trained neural network model, and comprising the following steps:
taking the trained convolution-pooling layer parameters as the convolution-pooling layer parameters of the neural network model, and fixing the convolution-pooling layer parameters of the neural network model;
the method comprises the steps of acquiring submarine positioning acquisition data, wherein the submarine positioning acquisition data are submarine target positioning information and acoustic wave information, expanding a full connection layer into three layers, and the full connection layer is as follows:
I＝f(w k D′+b k )
k＝1，2，3
wherein:
d' is input acoustic wave information, namely a pulse compression waveform result;
w k the weight of the k-th full-connection layer;
b k the offset of the k-th full-connection layer;
f (·) is the Relu activation function;
i represents the underwater target position information output by the neural network;
based on a training data set, training by utilizing an ADAM algorithm to obtain parameters of three full-connection layers, and fixing full-connection parameters of a first layer and a second layer, wherein the full-connection layer parameters comprise weights and offset of the full-connection layers;
the acoustic wave information in the submarine positioning acquisition data is used as the input of a neural network model, and parameters of a third full-connection layer are adjusted by comparing the underwater g-scale position information output by the neural network model with submarine target positioning information, so that the distance between the two position information is keptMinimum, where I h Indicating seabed target positioning information in seabed positioning acquisition data, < +.>The underwater target position information output by the neural network model is represented;
s5: the method comprises the steps of collecting actual underwater data, preprocessing the collected underwater data, taking the preprocessed underwater data as input of a neural network model, and outputting an output result as underwater target positioning coordinates.
2. The intelligent underwater target positioning method as claimed in claim 1, wherein the constructing an underwater target positioning mathematical model in step S1 comprises: constructing a signal receiving model, wherein the signal receiving model is as follows:
wherein:
S ij (t) when the ith signal transmitting array element transmits signals, the jth signal receiving array element receives signals, and t represents signal transmitting time;
c i (t) represents a signal transmitted by an ith signal transmitting element;
representing a convolution operation;
n (t) represents an underwater noise signal;
h ij (t) represents an impulse response function of the underwater signal channel;
(x, y) represents the two-dimensional coordinates of the underwater target, the underwater target receives L incident sound rays, the signal receiving array element receives Q scattered sound rays, and the impulse response of the incident sound rays p to the underwater target is G p (x, y) with a delay of T p P is more than or equal to 1 and less than or equal to L, and the impulse response of scattered sound rays reaching signal receiving array elements is G q (x, y) with a delay of T q ，1≤q≤Q；
Representing impulse response formed by L incident sound rays and Q scattered sound rays according to time delay T in time domain p +T q Superimposed together to form an underwater objectTarget impulse response function h' ij (t)。
3. The intelligent underwater target positioning method as claimed in claim 2, wherein the constructing the underwater target positioning mathematical model in step S1 comprises: according to the constructed signal receiving model, constructing an underwater target positioning mathematical model, and carrying out pulse compression processing on the received signals:
wherein:
* Representing complex conjugate;
D ij (t) represents the received signal S ij A pulse compression waveform result of (t);
constructing a mathematical model of a pulse compression waveform result and underwater target positioning coordinates:
wherein:
(x, y) represents the two-dimensional coordinates of the underwater target.
4. An intelligent underwater target positioning method as claimed in claim 3, wherein the generating of a large amount of test simulation data by using the constructed underwater target positioning mathematical model in the step S2 includes:
according to the constructed underwater target positioning mathematical model, simulating signal transmitting array elements, signal receiving array elements and an underwater environment, randomly generating a plurality of underwater targets, transmitting signals to the underwater targets by the simulating signal transmitting array elements, calculating an underwater impulse response function, taking the underwater impulse response function, the underwater targets and the transmitting signals as parameters of the underwater target positioning mathematical model, and obtaining an underwater target d k Corresponding pulse compression waveform result D k (t) where k.epsilon.1, E]E represents the generated underwater objectNumber of targets, underwater target d k Is positioned at the coordinate I k ＝(x k ，y k ) To be underwater target d k Position information I of (a) k Corresponding pulse compression waveform results D k (t) as a set of test simulation data, the test simulation data set is { (I) 1 ，D 1 (t))，(I 2 ，D 2 (t))，(I 3 ，D 3 (t))，...，(I E ，D E (t))}。
5. The intelligent underwater target positioning method according to claim 4, wherein the step S2 of preprocessing the generated test simulation data, using the preprocessed test simulation data as training data, comprises:
pre-emphasis pretreatment is carried out on pulse compression waveform results in a test simulation data set, wherein the pre-emphasis formula is as follows:
D′ i (t)＝D i (t)-aD i (t-1)
wherein:
a represents a pre-emphasis coefficient, which is set to 0.98;
D i (t) represents an underwater target d i The corresponding pulse compression waveform results;
D′ i (t) is the pulse compression waveform result after pre-emphasis treatment;
the test simulation data set after pretreatment { (I) 1 ，D′ 1 (t))，(I 2 ，D′ 2 (t))，...，(I E ，D′ E (t)) as a training data set.
6. The intelligent underwater target positioning method as claimed in claim 5, wherein the step S3 of inputting training data as input to the neural network model comprises:
inputting pulse compression waveform results in training data as a neural network model, outputting underwater target position information in the training data as the neural network model, and constructing the neural network model, wherein the loss function of the neural network model is as follows:
wherein:
e represents the number of underwater targets in the training data;
I e representing an underwater target d e Is a real position information of the mobile terminal;
underwater target d representing neural network model output e Is provided.
7. The intelligent underwater target positioning method according to claim 1, wherein in the step S5, underwater data of actual underwater target positioning is collected, the collected underwater data is preprocessed, the preprocessed underwater data is used as input of a neural network model, and an output result is positioning coordinates of the underwater target, and the method comprises the following steps:
the method comprises the steps of collecting underwater data of actual underwater target positioning, preprocessing the collected underwater data to obtain a pulse compression waveform result of the underwater data, taking the pulse compression waveform result as input of a neural network model, and outputting underwater target positioning coordinates by using the neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111351499.7A CN114216464B (en) | 2021-11-16 | 2021-11-16 | Intelligent underwater target positioning method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111351499.7A CN114216464B (en) | 2021-11-16 | 2021-11-16 | Intelligent underwater target positioning method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114216464A CN114216464A (en) | 2022-03-22 |
| CN114216464B true CN114216464B (en) | 2023-08-25 |
Family
ID=80697196
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111351499.7A Active CN114216464B (en) | 2021-11-16 | 2021-11-16 | Intelligent underwater target positioning method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114216464B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106093849A (en) * | 2016-06-03 | 2016-11-09 | 清华大学深圳研究生院 | A kind of Underwater Navigation method based on range finding with neural network algorithm |
| KR20190053470A (en) * | 2017-11-10 | 2019-05-20 | 주식회사 셀리지온 | Positioning system based on deep learnin and construction method thereof |
| CN111983619A (en) * | 2020-08-07 | 2020-11-24 | 西北工业大学 | Underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning |
| CN112556682A (en) * | 2021-02-07 | 2021-03-26 | 天津蓝鳍海洋工程有限公司 | Automatic target detection algorithm for underwater composite sensor |
| CN113406649A (en) * | 2021-06-21 | 2021-09-17 | 湖南国天电子科技有限公司 | Portable imaging sonar system for diver |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11769045B2 (en) * | 2018-12-27 | 2023-09-26 | Nokomis, Inc | Method to identify acoustic sources for anti-submarine warfare |
-
2021
- 2021-11-16 CN CN202111351499.7A patent/CN114216464B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106093849A (en) * | 2016-06-03 | 2016-11-09 | 清华大学深圳研究生院 | A kind of Underwater Navigation method based on range finding with neural network algorithm |
| KR20190053470A (en) * | 2017-11-10 | 2019-05-20 | 주식회사 셀리지온 | Positioning system based on deep learnin and construction method thereof |
| CN111983619A (en) * | 2020-08-07 | 2020-11-24 | 西北工业大学 | Underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning |
| CN112556682A (en) * | 2021-02-07 | 2021-03-26 | 天津蓝鳍海洋工程有限公司 | Automatic target detection algorithm for underwater composite sensor |
| CN113406649A (en) * | 2021-06-21 | 2021-09-17 | 湖南国天电子科技有限公司 | Portable imaging sonar system for diver |
Non-Patent Citations (1)
| Title |
|---|
| zou junhua 等.convolution neural network simplification via feature map pruning.computer and electrical engineering.2018,第70卷第950-958页. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114216464A (en) | 2022-03-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang et al. | Source localization using deep neural networks in a shallow water environment | |
| Abraham | Underwater Acoustic Signal Processing: Modeling, Detection, and Estimation | |
| Wahlberg et al. | Source parameters of echolocation clicks from wild bottlenose dolphins (Tursiops aduncus and Tursiops truncatus) | |
| CN103076594B (en) | Method for positioning underwater sound pulse signal by double array elements on basis of cross-correlation | |
| CN103076590A (en) | Method for positioning underwater sound pulse signal on basis of frequency estimation | |
| CN109541548A (en) | A kind of air sonar localization method based on Matched Field | |
| Yoon et al. | Deep learning-based high-frequency source depth estimation using a single sensor | |
| CN115169216A (en) | High-robustness confrontation sample generation method and system for underwater sound intelligent camouflage | |
| CN108572349B (en) | Sound source depth setting method based on model calculation under deep sea environment | |
| CN114216464B (en) | Intelligent underwater target positioning method | |
| CN112511469B (en) | Sparse underwater acoustic channel estimation method based on deep learning | |
| Barros et al. | A computational Bayesian approach for localizing an acoustic scatterer in a stratified ocean environment | |
| CN113109794B (en) | Deep sea sound source depth setting method based on deep neural network in strong noise environment | |
| CN107635181A (en) | A kind of multiple access based on channel study perceives the feedback optimized method in source | |
| La Cour et al. | Multi-everything sonar simulator (mess) | |
| Zamanizadeh et al. | Source localization from time-differences of arrival using high-frequency communication signals | |
| CN105372644A (en) | adaptive wave beam formation method based on dynamic re-correction and system | |
| CN113126029B (en) | Multi-sensor pulse sound source positioning method suitable for deep sea reliable acoustic path environment | |
| Goldwater et al. | Machine-learning-based simultaneous detection and ranging of impulsive baleen whale vocalizations using a single hydrophone | |
| CN114974298A (en) | Ocean sound source positioning and ocean parameter inversion method based on deep learning | |
| CN115220026A (en) | Matching field ice-crossing positioning method based on ice surface horizontal array | |
| Valtierra et al. | Calling depths of baleen whales from single sensor data: Development of an autocorrelation method using multipath localization | |
| Wazenski et al. | Active, wideband detection and localization in an uncertain multipath environment | |
| CN111965601A (en) | Underwater sound source passive positioning method based on nuclear extreme learning machine | |
| Andrieux et al. | Lake‐bottom recognition using a wideband sonar system and time‐frequency analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |