CN117454102A - Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA - Google Patents
Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA Download PDFInfo
- Publication number
- CN117454102A CN117454102A CN202311340149.XA CN202311340149A CN117454102A CN 117454102 A CN117454102 A CN 117454102A CN 202311340149 A CN202311340149 A CN 202311340149A CN 117454102 A CN117454102 A CN 117454102A
- Authority
- CN
- China
- Prior art keywords
- neural network
- signal
- designing
- network
- noise
- 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
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000001514 detection method Methods 0.000 title claims abstract description 41
- 230000008030 elimination Effects 0.000 title claims abstract description 23
- 238000003379 elimination reaction Methods 0.000 title claims abstract description 23
- 238000013528 artificial neural network Methods 0.000 claims abstract description 47
- 230000003044 adaptive effect Effects 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 25
- 230000008569 process Effects 0.000 claims abstract description 19
- 230000002068 genetic effect Effects 0.000 claims abstract description 16
- 238000013459 approach Methods 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 38
- 230000004913 activation Effects 0.000 claims description 10
- 108090000623 proteins and genes Proteins 0.000 claims description 10
- 230000035772 mutation Effects 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 5
- 210000002569 neuron Anatomy 0.000 claims description 5
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 claims description 2
- 238000006467 substitution reaction Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 238000009933 burial Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 3
- 230000002596 correlated effect Effects 0.000 description 2
- 239000000523 sample Substances 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003750 conditioning effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/17—Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method
-
- 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Mathematical Optimization (AREA)
- Computational Linguistics (AREA)
- Pure & Applied Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Genetics & Genomics (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Neurology (AREA)
- Filters That Use Time-Delay Elements (AREA)
Abstract
The method and the device for eliminating self-adaptive noise of the river-crossing pipeline positioning detection system based on the FPGA comprise the following steps: step one, designing an adaptive filter to primarily process an input signal. Step two, designing a neural network to further process the input signal to approach the useful signal. Training the radial basis function neural network by using an improved genetic algorithm, deducing a neural network structure based on the FPGA, and obtaining network optimal parameters. And step four, designing a hardware deployment scheme of the noise elimination system. And fifthly, denoising the useful signal containing noise by using the constructed noise cancellation system. The invention can reduce the noise elimination error of useful signals and improve the positioning and burial depth detection precision of the river crossing pipeline.
Description
Technical Field
The invention relates to the technical field of pipeline detection weak signal denoising, in particular to a self-adaptive noise elimination method and device of a river-crossing pipeline positioning detection system based on an FPGA.
Background
With the vigorous development of science and technology in recent years, pipeline detection technology based on an electromagnetic method is widely applied. Electromagnetic pipeline detection technology is used as one of pipeline detection technologies for long-term research, and the magnetic field intensity excited by a pipeline is detected by introducing a current signal with a certain frequency into the pipeline so as to acquire the information of the buried depth and the plane position of the pipeline,
however, as the depth of the river crossing pipeline increases, the surrounding environment is more complex, and in the positioning and detecting process of the river crossing pipeline, the acquired signals are weak and are easily influenced by the surrounding environment and the circuit of the river crossing pipeline, so that the pipeline detection is challenged.
Therefore, a method for reducing the noise cancellation error of the useful signal and improving the positioning and burial depth detection accuracy of the river crossing pipeline is needed.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides the self-adaptive noise elimination method and the self-adaptive noise elimination device for the river crossing pipeline positioning detection system based on the FPGA, which can reduce the noise elimination error of useful signals and improve the positioning and burial depth detection precision of the river crossing pipeline.
In order to achieve the purpose, the self-adaptive noise elimination method of the river crossing pipeline positioning detection system based on the FPGA comprises the following steps:
step one, designing an adaptive filter to primarily process an input signal.
Step two, designing a neural network to further process the input signal to approach the useful signal.
Training the radial basis function neural network by using an improved genetic algorithm, deducing a neural network structure based on the FPGA, and obtaining network optimal parameters.
And step four, designing a hardware deployment scheme of the noise elimination system.
And fifthly, denoising the useful signal containing noise by using the constructed noise cancellation system.
Further, the designing the adaptive filter in the first step primarily processes the input signal, specifically includes:
inputting the two paths of channel signals into an adaptive filter, and preprocessing the signals;
the two paths of channel signals are signals received by the vertical coil and the horizontal coil respectively, and the vertical coil is used as a noise collector of the horizontal coil;
and designing an adaptive filter by using a least mean square algorithm, wherein the least mean square value of the difference value between the input signal and the output signal of the filter is used as an optimal statistical criterion, and the error signal output by the filter is the primary estimation of the useful signal after the signal preprocessing.
Further, the designing neural network in the second step further processes the input signal to approach the useful signal, specifically includes:
inputting the error signal output by the adaptive filter into the constructed neural network, further eliminating irrelevant noise, and approaching to useful signals;
wherein the neural network is designed as a radial basis neural network using a gaussian function as the neural network activation function.
Further, training the radial basis function neural network by using the improved genetic algorithm, deducing a neural network structure based on the FPGA, and obtaining network optimal parameters, wherein the method specifically comprises the following steps:
the genetic algorithm is improved, which comprises the steps of designing a crossover operator and a mutation operator;
and training the radial basis function neural network by using the improved genetic algorithm, executing the gradient descent method for a plurality of times, determining the iteration execution times according to the effect of repeated training, and outputting the center vector, the width parameter and the weight of the radial basis function neural network.
Further, the hardware deployment scheme for designing the noise cancellation system in the fourth step specifically includes:
the first stage adaptive filter hardware is designed using a standard parallel adaptive mode.
The neural network hardware implementation method and the network training method adopt different schemes, namely, a genetic algorithm is adopted to train the central vector, the width parameters and the weight values of the network in the network training stage, the central vector and the width parameters which are trained in advance are stored into the RAM when the hardware is deployed, and the trained weight values are used as initial values of linear output combination.
And constructing a network hidden layer neuron activation function by using a CORDIC algorithm, and updating the system weight by using an LMS algorithm.
The second aspect of the invention relates to an adaptive noise cancellation device of an FPGA-based river crossing pipeline positioning detection system, comprising a memory and one or more processors, wherein executable codes are stored in the memory, and the one or more processors are used for realizing the adaptive noise cancellation method of the FPGA-based river crossing pipeline positioning detection system when the executable codes are executed.
A third aspect of the present invention relates to a computer readable storage medium having stored thereon a program which, when executed by a processor, implements the FPGA-based river crossing pipeline location detection system adaptive noise cancellation method of the present invention.
The beneficial effects of the invention are as follows:
according to the invention, the self-adaptive filter is designed and the genetic algorithm is improved to train the radial basis function neural network, so that the hardware deployment scheme of the noise elimination system is determined, the noise elimination error of useful signals can be reduced, and the positioning and burial depth detection precision of the pipeline crossing the river can be improved.
Drawings
Fig. 1 is a schematic diagram of a basic structure of a noise cancellation model according to an embodiment of the present invention.
Fig. 2 is a schematic diagram of an adaptive filter according to an embodiment of the present invention.
FIG. 3 is a flowchart of an adaptive filter algorithm in an embodiment of the present invention.
Fig. 4 is a schematic diagram of an RBF neural network according to an embodiment of the invention.
Fig. 5 is a schematic diagram of the overall structure of a noise cancellation model according to an embodiment of the present invention.
Fig. 6 is a schematic diagram of a correspondence relationship between a hardware deployment scenario and a detection flow in an embodiment of the present invention.
Fig. 7 is a flowchart of a method for eliminating adaptive noise of a river crossing pipeline positioning detection system based on an FPGA according to an embodiment of the present invention.
Detailed Description
The invention will now be described in detail by way of example with reference to the accompanying drawings.
Example 1
In this embodiment, the river-crossing pipeline positioning detection system adopts an electromagnetic method principle, uses a river-crossing cable to connect a pipeline into a closed loop, connects a transmitter in series to the pipeline loop, continuously transmits a low-frequency sine wave signal to a pipeline to be detected by the transmitter, acquires pipeline magnetic field information by using a T-shaped receiving probe, processes the acquired signal in a receiver, and finally converts the pipeline magnetic field signal into buried depth and position information of the river-crossing pipeline.
In the river crossing pipeline positioning detection process, the acquired signals are weak and are easily influenced by surrounding environment noise and self circuit noise.
Therefore, how to effectively eliminate noise and extract useful signals in positioning and detecting the crossing river pipeline is an effective way for improving the positioning and burial depth detection precision of the crossing river pipeline.
The self-adaptive noise elimination method of the river-crossing pipeline positioning detection system based on the FPGA provided by the embodiment of the invention has the main flow shown in figure 7, and comprises the following steps:
step one, designing an adaptive filter to primarily process an input signal.
By utilizing the principle of an electromagnetic method, a horizontal coil of a sensor of the pipeline detection system receives a horizontal component of a magnetic field signal of an underwater pipeline, a vertical coil receives a vertical component of the magnetic field signal, and two paths of signals are subjected to preliminary processing by a signal conditioning module and then are subjected to noise elimination by an analog-digital conversion input system.
And taking the signals received by the vertical coil as a noise collector of the horizontal coil, inputting the two paths of channel signals into the adaptive filter, and preprocessing the signals.
The basic structure of the adaptive filter is shown in fig. 2.
Where d (N) =x (N) +n (N) represents an input signal at time N, which is a superposition of a useful signal and noise, N '(N) represents another input signal of the filter at time N as a desire of the system, and N' (N) is strongly correlated with N (N) and weakly correlated with x (N).
y (n) represents the actual output signal of the filter at time n, is an estimate of the noise, e 1 (n) represents an error signal at time n, and is outputted in the system as an estimate of the useful signal x (n), and e is calculated by the equation (1) as a difference between the desired signal d (n) and the actual output signal y (n) 1 (n)。
e 1 (b)＝d(n)-y(n)＝x(n)+N(n)-y(n) (1)
The adaptive filter uses a least mean square algorithm to adjust the parameters of the filter to make the error signal e 1 The difference between (n) and the useful signal x (n) is minimized.
Let the input signal be X (n) = [ X (n), X (n-1), …, X (n-m)] T
The output y (n) of the adaptive filter is calculated by equation (2).
Wherein W (n) = [ W 0 (n),w 1 (n),…,w m (n)] T ，w m (n) is the weight coefficient of the filter, and m is the filter order.
The system takes the least mean square value of the difference value between the filter output signal y (N) and the input signal N' (N) as the optimal statistical criterion, and the filter cost function J 1 (n) is represented by formula (3).
J 1 (n)＝E[(y(n)-N’(n)) 2 ] (3)
The updating method of the filter coefficient w (n) is shown in the formula (4).
In the middle ofIn order to be along the gradient direction, mu is a filter step factor, and the value range is as follows:
λ max is the maximum eigenvalue of the input signal autocorrelation matrix W. Then:
the updated formula of the filter coefficient w (n) substituted into the formula (4) is shown in the formula (5).
w(n+1)＝w(n)+2μe(n)X(n) (5)
Because the step mu is fixed, the steady state error of the adaptive filter in the system and the convergence speed have contradiction that is difficult to coordinate, namely, the large step makes the steady state error smaller, but the convergence speed is slower; the small step length makes the convergence speed faster, but the error is larger in the steady state, and the parameters need to be adjusted to make the signal preprocessing achieve the ideal effect. The pipeline detection system is trained, the step size mu is 0.000986, the filter order is 4, and the algorithm flow chart of the adaptive filter is shown in fig. 3.
Step two, designing a neural network to further process the input signal to approach the useful signal.
After the noise of the input signal is preliminarily eliminated by the adaptive filter, the error signal e output by the adaptive filter is processed by adopting the idea of machine learning 1 (n) inputting the constructed neural network, further eliminating extraneous noise to approximate the useful signal, the neural network structure is shown in fig. 4.
Wherein, RBF neural network adopts LMS algorithm to carry out iterative update, calculates neural network output y through formula (6) 2 (n)。
Where m is the number of nodes in the hidden layer, ω ji For the weight from the hidden layer to the output layer of the network, h j (x) Is a neural netThe complex basis function, the system is defined as a Gaussian kernel function, and the Gaussian kernel function is shown in a formula (7).
In sigma j Representing the width of the gaussian kernel, ||x-c j ‖ 2 The euclidean norms of the input parameter vector x and the hidden layer node center vector c are represented.
Output y in a neural network in an RBF nonlinear processing system 2 (n) and the previous stage output e 1 (n) the mean square value of the difference as a cost function, thereby making RBF output y 2 (n) ideally approximates the useful signal x (n), the cost function J being calculated by equation (8) 2 (n)。
J 2 (n)＝E[(y 2 (n)-e 1 (n)) 2 ] (8)
The self-adaptive noise elimination method of the underwater pipeline detection system consists of a self-adaptive filtering system and an RBF neural network, the signal to noise ratio of the useful signal is initially improved after the input signal is subjected to noise elimination through the self-adaptive filter, and the useful signal is approximated through the radial basis neural network, wherein the composition structure of the self-adaptive noise elimination method is shown in figure 1.
Training the radial basis function neural network by using an improved genetic algorithm, deducing a neural network structure based on the FPGA, and obtaining network optimal parameters.
Biological genes were simulated using binary 01 sequences, based on precision pre and interval sec [ l ] dem ,r dem ]Determining the length of a gene sequence, establishing a binary-decimal-independent variable mapping relation, establishing a population gene library by a random operator, and outputting a value x in an independent variable interval as an iterative initial population. Generating gene sequence indiv, calculating length of gene sequence The mapping relationship between the gene and the independent variable is represented by formula (9).
Taking an independent variable x generated by a random operator as the center, width parameters and weight of RBF neural network nodes to participate in evolution, and outputting a signal y from a system 2 (n) selecting a next generation population as a fitness value, the fitness value and the systematic error signal e 2 (n) corresponds to.
Smaller errors are selected as individuals with greater fitness to participate in the next evolution.
(1) Crossover operator
The strategy of gene exchange among individuals is adopted, a certain cross probability is set, partial individuals which are not selected by a selection operator are randomly selected for cross inheritance, the individuals after cross join in a population to participate in the next evolution, and the specific algorithm is as follows:
(2) Mutation operator
The strategy of 0-1 replacement of the self genes of individuals is adopted, a certain mutation probability is set, partial individuals of the population after the last evolution are randomly selected for mutation operation, the individuals after mutation are added into the population to participate in the next evolution, and the specific algorithm is as follows:
pre-training the generated initial parameters by gradient descent method, and learning the network basis function center c j (n) and Width parameter sigma j (n) determining the execution times according to the effect of the repeated training, and weighting omega from the hidden layer to the output layer for the network ji Training is performed by the formula (10), the formula (11), and the formula (12), and the adjustment amounts of the parameters are calculated.
U in the formula ω ，u c ，u σ For the learning rate of each parameter in the iterative process, the value range is
In the training process, the accuracy cannot be ensured by too few network layers, the detection instantaneity can be influenced by too high hidden layers, and meanwhile, the resource problem and difficulty degree of hardware realization are considered, so that the system is designed to be an N-4-1 type network.
And step four, designing a hardware deployment scheme of the noise elimination system.
The first stage adaptive filter hardware design employs a standard parallel adaptive mode.
The neural network hardware implementation method and the network training method adopt different schemes, namely, a genetic algorithm is adopted to train the central vector, the width parameters and the weight values of the network in the network training stage, the central vector and the width parameters which are trained in advance are stored into the RAM when the hardware is deployed, and the trained weight values are used as initial values of linear output combination.
And constructing a network hidden layer neuron activation function by using a CORDIC algorithm, and updating the system weight by using an LMS algorithm.
The CORDIC core selects a hyperbolic function conversion function, and the conversion relation between the activation function and the double Qu Zheng cosine function is shown in formula (13).
e θ ＝sinhθ+coshθ (13)
The data processing process is divided into two stages, namely a training stage and an actual detection stage. In the data training process, the system network training stage corresponds to a calibration point elevation test in actual detection, and is used as a fitting curve interpolation standard of detection points on the water surface, and signals received by two paths of channels in a receiving probe are equivalent to a test set; in the actual detection process, the hardware implementation stage corresponds to the water surface magnetic field signal acquisition in detection, the designed noise elimination system is utilized for digital signal processing, signals received by the two paths of channels are equivalent to a test set, and the corresponding relation between the second-stage signal processing hardware deployment scheme and the detection flow is shown in fig. 6.
And fifthly, denoising the useful signal containing noise by using the constructed noise cancellation system.
Example 2
The embodiment relates to an adaptive noise elimination device of a river crossing pipeline positioning detection system based on an FPGA, which comprises a memory and one or more processors, wherein executable codes are stored in the memory, and the one or more processors are used for realizing the adaptive noise elimination method of the river crossing pipeline positioning detection system based on the FPGA in the embodiment 1 when executing the executable codes.
Example 3
The present embodiment relates to a computer-readable storage medium having stored thereon a program which, when executed by a processor, implements the adaptive noise canceling method of the FPGA-based river crossing pipeline positioning detection system of embodiment 1.
The foregoing detailed description of the embodiments of the invention is merely a preferred embodiment of the invention and is not intended to limit the scope of the invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are within the scope of the present invention.
Claims (10)
1. The adaptive noise elimination method of the river-crossing pipeline positioning detection system based on the FPGA is characterized by comprising the following steps of:
firstly, designing an adaptive filter to primarily process an input signal;
step two, designing a neural network to further process the input signal to approach the useful signal;
training a radial basis neural network by using an improved genetic algorithm, deducing a neural network structure based on the FPGA, and obtaining network optimal parameters;
designing a hardware deployment scheme of the noise elimination system;
and fifthly, denoising the useful signal containing noise by using the constructed noise cancellation system.
2. The method of claim 1, wherein the designing the adaptive filter in step one initially processes the input signal, specifically comprises:
inputting the two paths of channel signals into an adaptive filter, and preprocessing the signals;
the two paths of channel signals are signals received by the vertical coil and the horizontal coil respectively, and the vertical coil is used as a noise collector of the horizontal coil;
and designing an adaptive filter by using a least mean square algorithm, wherein an error signal output by the filter is a useful signal preliminary estimation after signal preprocessing.
3. The method of claim 2, wherein said adaptive filter d (N) represents an input signal at time N, is a superposition of the desired signal x (N) and noise N (N), N' (N) represents another input signal of the time N filter, y (N) represents an actual output signal of the time N filter, e 1 (n) represents an error signal at time n:
e 1 (n)＝d(n)-y(n)＝x(n)+N(n)-y(n) (1)
let the input signal be X (n) = [ X (n), X (n-1), …, X (n-m)] T Weights of filtersCoefficient W (n) = [ W ] 0 (n),w 1 (n),…,w m (n)] T The calculation formula of the output signal of the adaptive filter is:
the least mean square value of the difference between the input signal and the output signal of the filter is used as the optimal statistical criterion to calculate the cost function J 1 (n)：
J 1 (n)＝E[(y(n)-N’(n)) 2 ] (3)
The update formula of the filter coefficient w (n) is as follows:
wherein the method comprises the steps ofIn order to be along the gradient direction, mu is a filter step factor, and the value range is as follows: />λ max For the maximum eigenvalue of the autocorrelation matrix w of the input signal,/>Substitution into equation (4) yields an updated equation for the filter coefficient w (n):
w(n+1)＝w(n)+2μe(n)X(n) (5)
4. the method of claim 1, wherein the designing the neural network in step two further processes the input signal to approximate the desired signal, and specifically comprises:
inputting the error signal output by the adaptive filter into the constructed neural network, further eliminating irrelevant noise, and approaching to useful signals;
the designed neural network is a radial basis function neural network, and output signals of the neural network are calculated:
where m is the number of nodes in the hidden layer, ω ji For the weight from the hidden layer to the output layer of the network, h j (x) For the neural network activation function, a gaussian function is used as the neural network activation function:
wherein sigma j Representing the width of the gaussian kernel, ||x-c j ‖ 2 Euclidean norms representing the input parameter vector x and the hidden layer node center vector c, and the neural network cost function is the network output y 2 (n) and adaptive filter output e 1 (n) mean square function of the difference:
J 2 (n)＝E[(y 2 (n)-e 1 (n)) 2 ] (8)
5. the method of claim 1, wherein the training the radial basis function neural network using the improved genetic algorithm in the third step derives a neural network structure based on the FPGA, and obtains the network optimal parameters, specifically including:
the genetic algorithm is improved, which comprises the steps of designing a crossover operator and a mutation operator;
and training the radial basis function neural network by using the improved genetic algorithm, executing the gradient descent method for a plurality of times, determining the iteration execution times according to the effect of repeated training, and outputting the center vector, the width parameter and the weight of the radial basis function neural network.
6. The method of claim 5, wherein improving the genetic algorithm comprises:
according to the precision pre and interval sec [ l ] dem ,r dem ]Determination of the Gene sequence LengthThe mapping relation of the gene-independent variable is as follows:
training a radial basis neural network by using an improved genetic algorithm, executing a gradient descent method for a plurality of times, determining iteration execution times according to the effect of repeated training, and calculating a parameter adjustment quantity delta omega ji (b)、Δc j (n)、Δσ j (n)：
Wherein u is ω ，u c ，u σ For the learning rate of each parameter in the iterative process, the value range is
7. The method of claim 1, wherein the designing the hardware deployment of the noise cancellation system in step four specifically comprises:
designing first-stage adaptive filter hardware by adopting a standard parallel adaptive mode;
the neural network hardware implementation method and the network training method adopt different schemes, namely, a genetic algorithm is adopted to train a central vector, width parameters and weights of a network in a network training stage, the central vector and the width parameters which are trained in advance are stored into a RAM when hardware is deployed, and the trained weights are used as initial values of linear output combination;
and constructing a network hidden layer neuron activation function by using a CORDIC algorithm, and updating the system weight by using an LMS algorithm.
8. The method of claim 7, wherein constructing a network hidden layer neuron activation function using a CORDIC algorithm and updating system weights using an LMS algorithm, specifically comprises:
updating the system weight by using an LMS algorithm, constructing a network hidden layer neuron activation function by using a CORDIC algorithm, wherein the CORDIC selects a hyperbolic function conversion function, and the activation function has the following conversion relation with double Qu Zheng and cosine functions:
e θ ＝sinhθ+coshθ (13)
9. an adaptive noise cancellation device for an FPGA-based river-crossing pipeline positioning detection system, comprising a memory and one or more processors, wherein the memory stores executable code, and the one or more processors are configured to implement the FPGA-based river-crossing pipeline positioning detection system adaptive noise cancellation method of any one of claims 1-8 when executing the executable code.
10. A computer readable storage medium having stored thereon a program which, when executed by a processor, implements the FPGA-based river crossing pipeline location detection system adaptive noise cancellation method of any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311340149.XA CN117454102A (en) | 2023-10-16 | 2023-10-16 | Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311340149.XA CN117454102A (en) | 2023-10-16 | 2023-10-16 | Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117454102A true CN117454102A (en) | 2024-01-26 |
Family
ID=89592116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311340149.XA Pending CN117454102A (en) | 2023-10-16 | 2023-10-16 | Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117454102A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118074675A (en) * | 2024-04-25 | 2024-05-24 | 深圳市鲸视科技有限公司 | Adaptive filtering method, device, computer equipment and storage medium |
-
2023
- 2023-10-16 CN CN202311340149.XA patent/CN117454102A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118074675A (en) * | 2024-04-25 | 2024-05-24 | 深圳市鲸视科技有限公司 | Adaptive filtering method, device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109993280B (en) | Underwater sound source positioning method based on deep learning | |
| CN114595732B (en) | Radar radiation source sorting method based on depth clustering | |
| CN110007265A (en) | A kind of Wave arrival direction estimating method based on deep neural network | |
| CN113864665B (en) | Fluid pipeline leakage positioning method based on adaptive ICA and improved RLS filter | |
| CN117454102A (en) | Self-adaptive noise elimination method and device for river-crossing pipeline positioning detection system based on FPGA | |
| CN107145943A (en) | Method based on the detection echo state network small-signal for improving teaching optimized algorithm in Chaotic Background | |
| CN107832789B (en) | Feature weighting K nearest neighbor fault diagnosis method based on average influence value data transformation | |
| CN108805287A (en) | The Gyro Random error compensating method and system of genetic wavelet neural network | |
| CN112711032A (en) | Radar target detection method and system based on graph data and GCN | |
| CN116125386A (en) | Intelligent positioning method and system for underwater vehicle with enhanced sparse underwater acoustic ranging | |
| CN110852440A (en) | Ocean front detection method based on dynamic fuzzy neural network | |
| CN114814776A (en) | PD radar target detection method based on graph attention network and transfer learning | |
| CN113642591B (en) | Multi-beam submarine sediment layer type estimation method and system based on transfer learning | |
| CN110852451B (en) | Recursive kernel self-adaptive filtering method based on kernel function | |
| CN115860113B (en) | Training method and related device for self-countermeasure neural network model | |
| CN116418633A (en) | Depth expansion underwater sound channel estimation method based on sparse and low-rank characteristics | |
| CN109217844B (en) | Hyper-parameter optimization method based on pre-training random Fourier feature kernel LMS | |
| CN115047408B (en) | Underwater multi-sound-source positioning method based on single-layer large convolution kernel neural network | |
| CN117521335A (en) | Acoustic velocity profile inversion method based on array multi-path delay structure | |
| JP2021028816A (en) | Computer system and model learning method | |
| CN113269309B (en) | RBF neural network improved UKF method suitable for weak signal extraction | |
| CN115015831B (en) | Large-scale array target azimuth estimation method under combined influence of impulse noise and non-uniform noise | |
| CN114219088B (en) | Decomposition method of aliasing shock wave signal | |
| CN111898476A (en) | Self-adaptive line spectrum enhancement method for coupling stochastic resonance | |
| CN117974736B (en) | Underwater sensor output signal noise reduction method and system based on machine learning |
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 |