High-resolution positioning method for noise source imaging

Abstract

The invention provides a high-resolution positioning method for noise source imaging. The high-resolution positioning method comprises the following steps: designing a sensor array layout and selecting sensors; a multi-channel acquisition synchronous acquisition sensor and records signals to obtain acquisition data; performing signal processing on the acquired data by using adaptive filtering, machine learning calibration, time-frequency analysis and intelligent beam forming technologies to obtain standard data; based on the standard data, performing high-resolution direction estimation by using an optimized multi-signal classification algorithm, and removing a sidelobe effect by using an optimized CLEAN algorithm to complete processing of the standard data; and based on the processed standard data, reconstructing a spatial distribution image of the noise source by using an inversion technology, and dynamically adjusting the resolution of the reconstructed spatial distribution image by using an adaptive algorithm to realize high-resolution positioning of noise source imaging. According to the invention, high-resolution positioning can be realized, the reliability and precision of noise detection are improved, and noise control and management are more effective.

Description

Technical Field

[0001] The invention relates to the technical field of noise source imaging, and in particular to a high-resolution positioning method for noise source imaging. Background Art

[0002] Noise source imaging is a method used to identify and locate noise sources in acoustic, mechanical or electronic systems. This technology is extremely important for engineers and researchers because it helps them find out the causes of noise and take effective measures to reduce or eliminate it. This is of great significance in areas such as environmental noise control, product design optimization, and equipment maintenance. However, with the acceleration of industrialization, the widespread use of mechanical equipment in actual industrial needs has brought significant noise pollution problems. Noise not only affects human health and comfort, but also affects the ecological environment. It can be seen that noise source imaging technology is of great significance for environmental protection.

[0003] Existing noise source imaging methods include: Acoustic Camera: Using a microphone array to capture sound pressure data and generate an image of the noise source. Beamforming: A signal processing technique for spatial filtering and noise source localization. Near-field Acoustic Holography: Reconstructing the sound field by taking acoustic measurements in the near field. Acoustic inverse problem: Locating the noise source by performing computational inversion. And so on.

[0004] In noise source imaging, resolution positioning is very important. Higher resolution allows for the identification of more subtle or closer noise sources, which is critical for accurate diagnosis of complex systems. At the same time, it can also help engineers to more accurately control noise sources and optimize the design of systems, improve the reliability and accuracy of noise source identification, and provide a strong foundation for environmental protection and algorithm simplification. However, the existing technologies currently have resolution positioning defects. Traditional methods such as beamforming technology are limited in resolution by the array aperture and signal frequency, making it difficult to distinguish adjacent noise sources. At the same time, the sidelobe effect of existing methods may lead to the emergence of false noise sources, affecting the accuracy of imaging. Therefore, the present invention proposes a high-resolution positioning method for noise source imaging to solve the above problems. Summary of the invention

[0005] In order to overcome the deficiencies of the prior art, the object of the present invention is to provide a high-resolution positioning method for noise source imaging, which can achieve high-resolution positioning, improve the reliability and accuracy of noise detection, and make noise control and management more effective.

[0006] To achieve the above object, the present invention provides the following solution: a high-resolution positioning method for noise source imaging, comprising the following steps:

[0007] Design sensor array layout, optimize array parameters, and select sensors;

[0008] Use high-precision clocks to synchronize all sensors and perform synchronous acquisition, then record the signals of each sensor through multi-channel acquisition to obtain collected data;

[0009] Performing signal processing on the collected data using adaptive filtering, machine learning calibration, time-frequency analysis, and intelligent beamforming technology to obtain standard data;

[0010] Based on the standard data, high-resolution direction estimation is performed using an optimized multiple signal classification algorithm, and sidelobe effects are removed using an optimized CLEAN algorithm to complete the processing of the standard data;

[0011] Based on the processed standard data, the spatial distribution image of the noise source is reconstructed using the inversion technology, and then the resolution of the reconstructed spatial distribution image is dynamically adjusted using an adaptive algorithm to achieve high-resolution positioning of the noise source imaging.

[0012] Optionally, design the sensor array layout and select sensors with high sensitivity and wide bandwidth, including:

[0013] Selecting an adaptive sensor array type according to the spatial dimension of the noise source location, and performing irregular distribution of sensor positions based on the selected sensor array type to obtain a sensor array;

[0014] Increasing the array aperture of the sensor array and selecting and setting the sensor spacing to complete array parameter optimization;

[0015] Based on the optimized sensor array, sensors with high sensitivity and wide bandwidth are selected, and the sensors are calibrated in amplitude and phase to achieve signal consistency and accuracy.

[0016] Optionally, a high-precision clock is used to synchronize all sensors and perform synchronous acquisition, and then the signals of each sensor are recorded by multi-channel acquisition to obtain the acquired data, including:

[0017] Select a sensor clock based on the GPS system, synchronize the sensor clock using the NTP protocol, and then calibrate the sensor clock regularly. Add a timestamp to the collected data of each sensor and broadcast a synchronization signal to complete the design of the synchronization mechanism and obtain the high-precision clock.

[0018] Design a multi-channel data acquisition system for simultaneously recording multiple sensor signals, and introduce a machine learning technique for adaptively adjusting the sampling rate and data acquisition strategy in the multi-channel data acquisition system;

[0019] Select high-speed data transmission protocols for real-time data transmission, and introduce edge computing technology in sensor nodes to reduce data transmission volume and improve response speed;

[0020] Data redundancy and backup strategies as well as data verification mechanisms are introduced to perform real-time processing of transmitted data to obtain collected data.

[0021] Optionally, the collected data is processed by using adaptive filtering, machine learning calibration, time-frequency analysis and intelligent beamforming technology to obtain standard data, including:

[0022] An adaptive filter for dynamically adjusting the frequency response is selected, a multi-stage filtering strategy is designed in the adaptive filter and a nonlinear filtering technology for processing non-Gaussian noise is integrated to remove irrelevant frequency components; the multi-stage filtering strategy includes a broadband filter for removing external interference and a narrowband filter for refined processing;

[0023] Designing an automatic calibration mechanism in which known frequency and amplitude are used as reference signals and a machine learning algorithm is used to predict and compensate for dynamically changing system errors to complete amplitude and phase calibration of sensor signals;

[0024] Combine overlapping window weighting technology to extract the time-frequency characteristics of the signal, and use STFT algorithm or wavelet transform to analyze the time-frequency characteristics of the signal;

[0025] The beam direction is dynamically adjusted using an adaptive beamforming algorithm, and a deep learning model for identifying the optimal direction and a multi-mode beam technology for enhancing signals in multiple directions are introduced into the adaptive beamforming algorithm to complete intelligent beamforming and obtain standard data.

[0026] Optionally, high-resolution direction estimation using an optimized multiple signal classification algorithm, including:

[0027] Integrate the standard data into a data matrix, perform block processing and covariance matrix calculation on the data matrix, and obtain a block covariance matrix;

[0028] A randomized matrix decomposition method and a GPU algorithm are introduced to perform eigenvalue decomposition on the block covariance matrix to obtain a signal subspace and a noise subspace. The eigenvector corresponding to the maximum eigenvalue is selected through a dimensionality reduction technique to construct a signal subspace, and the remaining eigenvectors are used to construct a noise subspace.

[0029] Construct a MUSIC spectrum function, scan possible noise source directions and calculate spectrum function values, and identify the peak position used to represent the noise source direction in the spectrum function values ​​in combination with a trained machine learning model;

[0030] The Bayesian statistical method is used to analyze the peak position and estimate the direction of the noise source.

[0031] Optionally, use the optimized CLEAN algorithm to remove sidelobe effects, including:

[0032] S1. Use the multi-band MUSIC algorithm to combine the information of different frequency bands to generate the initial noise source imaging;

[0033] S2. Design and train a machine learning model for automatically identifying and classifying main lobe signals and side lobe signals, perform adaptive main lobe position identification of the strongest signal in the initial noise source imaging, obtain the main lobe signal, and use statistical methods to verify the identification result in real time;

[0034] S3, in the initial noise source imaging, removing the contribution of the main lobe signal according to the real shape of the identified main lobe, and updating the remaining signal;

[0035] S4, repeating iterative steps S2-S3 until the remaining signal is lower than a preset threshold; wherein an adaptive control strategy is introduced in the iterative process to dynamically adjust the number of iterations, iteration step size and preset threshold according to signal changes;

[0036] S5. Superimpose all the removed main lobe signals and reconstruct the initial noise source imaging in combination with a multi-model projection technique to complete the removal of the side lobe effect.

[0037] Optionally, an inversion technique is used to reconstruct a spatial distribution image of the noise source, including:

[0038] A linear model of the noise source is established, and the linear model is expanded into a nonlinear model according to the actual noise situation; the linear equation of the linear model is expressed as:

[0039] Ax=b

[0040] Where A is the system matrix, x is the noise source distribution to be solved, and b is the measured data;

[0041] Using weighted least squares method combined with matrix decomposition technology and manifold learning technology, the weights are dynamically adjusted according to signal quality or signal-to-noise ratio to calculate the linear equations and minimize the calculation error;

[0042] A regularization term is introduced into the linear model, and an adaptive regularization strategy for dynamically adjusting the regularization parameter is implemented according to the regularization term to deal with pathological problems or noise influences and complete the reconstruction of the spatial distribution image; the expression of the regularization term is:

[0043] min||Ax-b|| 2 +λLx|| 2

[0044] Among them, ||Ax-b|| 2 is the minimization error of the linear equation system, λ is the regularization parameter, L is the regularization matrix, and x is the noise source distribution to be solved;

[0045] The contrastive learning method is used to improve the reconstruction quality by analyzing and comparing different reconstruction results.

[0046] Optionally, an adaptive algorithm is used to dynamically adjust the resolution of the reconstructed spatial distribution image to achieve high-resolution positioning of noise source imaging, including:

[0047] Automatically selecting an initial resolution that meets the standard data characteristics using a machine learning algorithm;

[0048] Using a convolutional neural network to identify detail areas in the reconstructed spatial distribution image, and combining signal characteristics and noise levels to determine areas in the detail areas that need to be enhanced;

[0049] Introducing a local adaptive high-resolution reconstruction technique in the detail area to customize the local resolution and maintain or reduce the resolution in the flat area;

[0050] Adopting adaptive iterative algorithm combined with online learning mechanism to dynamically adjust the resolution and image reconstruction parameters of different areas;

[0051] A generative adversarial network is used to synthesize regions of different resolutions into a final image, and adaptive noise suppression and super-resolution techniques are integrated to post-process the final image to eliminate artifacts caused by asynchronous stitching and complete high-resolution positioning of noise source imaging.

[0052] The present invention discloses the following technical effects by providing a high-resolution positioning method for noise source imaging:

[0053] 1. Consistency and integrity of sensor data time: Through the design and implementation of high-precision clock synchronization and multi-channel data acquisition systems, the time consistency and integrity of data during noise source imaging can be ensured. This is the basis for achieving high-resolution positioning and can significantly improve the accuracy and reliability of noise source imaging.

[0054] 2. Signal processing enhancement: 1) Adaptive filters (such as bandpass filters) can be used to remove irrelevant frequency components and obvious external interference, dynamically adjust the frequency response, and adapt to the changing environment and noise characteristics. 2) The amplitude and phase of the sensor signal are calibrated through the automatic calibration mechanism to improve the calibration accuracy. 3) The short-time Fourier transform (STFT) or wavelet transform is used to analyze the time-frequency characteristics of the signal, which can perform multi-granularity analysis on complex signals and improve the processing speed. 4) The use of intelligent beamforming technology can enhance signals in specific directions. The above method enables the present invention to accurately locate noise sources in a changeable and complex environment, which not only enhances the effect of signal processing, but also improves the flexibility and adaptability of the algorithm, thereby better meeting the needs of practical applications.

[0055] 3. Algorithm optimization to improve computing efficiency and accuracy: 1) The use of the optimized multiple signal classification (MUSIC) algorithm can achieve faster computing speed and higher positioning accuracy when dealing with complex noise source environments, while also being able to adapt to changing signal environments and larger sensor systems to achieve high-resolution direction estimation. 2) The optimized CLEAN algorithm is more robust and flexible, and can effectively remove sidelobe effects, thereby improving the clarity and accuracy of noise source imaging. At the same time, it can also intelligently handle the uncertainty and dynamics in the algorithm, providing stronger support for accurate imaging in complex signal environments.

[0056] 4. High-resolution noise source imaging: 1) The spatial distribution image of the noise source is reconstructed using optimized inversion technology, which greatly improves the accuracy and efficiency of the noise source imaging inversion process, reduces the computational burden and improves adaptability. 2) Introducing deep learning technology and intelligent algorithms in the dynamic adjustment process of resolution can dynamically and efficiently optimize the results of noise source imaging, improve the image detail performance, and achieve high-resolution positioning of noise source imaging.

[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0059] Figure 1 A schematic diagram of a method flow chart provided by an embodiment of the present invention;

[0060] Figure 2A schematic diagram of a signal processing flow provided by an embodiment of the present invention;

[0061] Figure 3 A schematic diagram of a high-resolution imaging process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0064] like Figure 1-3 As shown, the present invention provides a high-resolution positioning method for noise source imaging, comprising the following steps:

[0065] 1. Equipment layout

[0066] Design the sensor array layout and select sensors with high sensitivity and wide bandwidth; including:

[0067] 1.1 Array Design

[0068] An adaptive sensor array type is selected according to the spatial dimension of the noise source positioning, and the sensor positions are irregularly distributed based on the selected sensor array type to obtain a sensor array.

[0069] For example, uniform linear arrays are suitable for locating noise sources in one-dimensional space, and uniform planar arrays are suitable for locating in two-dimensional space. Commonly used irregular distributions include random arrays and spiral arrays.

[0070] 1.2 Array parameter optimization

[0071] The array aperture of the sensor array is increased (resolution can be improved, but the number of sensors and cost need to be weighed), and the sensor spacing is selected to avoid spatial aliasing, and the spacing should usually be less than half the signal wavelength to complete the array parameter optimization.

[0072] The appropriate array shape (such as circular, rectangular, L-shaped) can also be selected according to the application scenario to optimize the coverage and resolution.

[0073] 1.3 Sensor Selection

[0074] Based on the optimized sensor array, sensors with high sensitivity (ensuring that weak noise signals can be detected) and wide bandwidth (capturing noise signals in different frequency ranges) are selected, and the sensors are calibrated in amplitude and phase to achieve signal consistency and accuracy.

[0075] By carefully designing the sensor array layout and selecting appropriate sensors, the spatial sampling rate and resolution of noise source imaging can be significantly improved, thereby achieving high-precision noise source localization.

[0076] 2. Data Collection

[0077] Use high-precision clocks to synchronize all sensors and perform synchronous acquisition, then record the signals of each sensor through multi-channel acquisition to obtain the collected data; including:

[0078] 2.1 Time Synchronization

[0079] Select a sensor clock based on the GPS system (i.e., GPS clock, which is classified by time protocol into NTP / SNTP, PTP, and IRIG-B, etc., and available high-precision clock chips such as RX-8803SA, RX-4803SA, etc.), and use the Network Time Protocol (NTP) protocol to synchronize the sensor clock (ensuring millisecond-level synchronization accuracy), and then regularly calibrate the sensor clock (to compensate for clock drift caused by temperature changes and aging), add an accurate timestamp to the collected data of each sensor and broadcast a synchronization signal, complete the design of the synchronization mechanism, and obtain the high-precision clock.

[0080] 2.2 Data acquisition system design

[0081] A multi-channel data acquisition system for simultaneously recording multiple sensor signals is designed, and a machine learning technique for adaptively adjusting sampling rate and data acquisition strategy is introduced into the multi-channel data acquisition system to dynamically optimize data acquisition according to environmental changes and signal characteristics.

[0082] 2.3 Data Completeness and Accuracy

[0083] Select high-speed data transmission protocols (such as fiber optic communication or high-speed Ethernet) for real-time data transmission, and introduce edge computing technology in sensor nodes to reduce data transmission volume and improve response speed;

[0084] Data redundancy and backup strategies (to prevent data loss or damage) and data verification mechanisms (to detect and correct errors that may occur during transmission) are introduced to perform real-time processing of transmitted data and obtain collected data.

[0085] 3. Signal processing

[0086] like Figure 2As shown, the collected data is processed by using adaptive filtering, machine learning calibration, time-frequency analysis and intelligent beamforming technology to obtain standard data; including:

[0087] 3.1 Use adaptive filters to remove irrelevant frequency components

[0088] An adaptive filter for dynamically adjusting the frequency response is selected, a multi-stage filtering strategy is designed in the adaptive filter and a nonlinear filtering technology for processing non-Gaussian noise is integrated (to enhance the purity of the signal) to remove irrelevant frequency components; the multi-stage filtering strategy includes a broadband filter for removing external interference and a narrowband filter for refined processing.

[0089] 3.2 Amplitude and phase calibration of sensor signals

[0090] An automatic calibration mechanism is designed, in which known frequency and amplitude are used as reference signals, and a machine learning algorithm is used to predict and compensate for dynamically changing system errors to complete the amplitude and phase calibration of the sensor signal.

[0091] 3.3 Time-frequency feature analysis

[0092] Combine overlapping window weighting technology to extract the time-frequency characteristics of the signal, and use the STFT algorithm or wavelet transform to analyze the time-frequency characteristics of the signal. Among them, the fast STFT algorithm is used, and the window optimization technology is used to reduce the calculation burden and improve the processing speed. The wavelet packet transform is used instead of the traditional wavelet transform to provide higher frequency resolution and perform multi-granularity analysis on complex signals.

[0093] 3.4 Signal Enhancement

[0094] The beam direction is dynamically adjusted using an adaptive beamforming algorithm, and a deep learning model (such as a convolutional neural network) for identifying the optimal direction and a multi-mode beam technology for enhancing signals in multiple directions are introduced into the adaptive beamforming algorithm to complete intelligent beamforming and obtain standard data.

[0095] Combining the above steps makes it possible to accurately locate noise sources in a changing and complex environment. These methods not only enhance the effect of signal processing, but also improve the flexibility and adaptability of the algorithm, thereby better meeting the needs of practical applications.

[0096] 4. Algorithm optimization

[0097] like Figure 3 As shown, based on the standard data, an optimized multiple signal classification algorithm is used to perform high-resolution direction estimation, and an optimized CLEAN algorithm is used to remove sidelobe effects to complete the processing of the standard data.

[0098] Among them, MUSIC (Multiple Signal Classification) is a high-resolution direction estimation algorithm based on eigenvalue decomposition. It uses the orthogonality of the signal and noise subspaces to identify the direction of the signal. The CLEAN algorithm is an iterative deconvolution algorithm that was originally used to remove sidelobe effects in radio astronomy. It can now reduce the impact of sidelobes by gradually removing the contribution of the mainlobe signal.

[0099] 4.1 High-resolution direction estimation using optimized multiple signal classification algorithm:

[0100] Integrate the standard data into a data matrix, perform block processing and covariance matrix calculation on the data matrix, and obtain a block covariance matrix to reduce storage requirements and calculation complexity, and is also applicable to large-scale sensor arrays;

[0101] A randomized matrix decomposition method and a GPU algorithm (which can accelerate eigenvalue decomposition and spectral peak search) are introduced to perform eigenvalue decomposition on the block covariance matrix. For example, randomized SVD (singular value decomposition) can significantly reduce the calculation time to obtain a signal subspace and a noise subspace.

[0102] The eigenvector corresponding to the maximum eigenvalue is selected through dimensionality reduction techniques (such as PCA) to construct the signal subspace, and the remaining eigenvectors are used to construct the noise subspace;

[0103] Construct a MUSIC spectrum function, scan possible noise source directions and calculate the spectrum function value, and use the trained machine learning model to identify the peak position used to represent the noise source direction in the spectrum function value, thereby improving the accuracy of peak recognition;

[0104] The Bayesian statistical method is used to analyze the peak position and estimate the direction of the noise source.

[0105] Through the above steps, the MUSIC algorithm can achieve faster calculation speed and higher positioning accuracy when dealing with complex noise source environments. At the same time, it also makes the MUSIC algorithm more adaptable to changing signal environments and larger-scale sensor systems.

[0106] 4.2 Use the optimized CLEAN algorithm to remove the sidelobe effect:

[0107] S1. Use the multi-band MUSIC algorithm to combine the information of different frequency bands to generate the initial noise source imaging;

[0108] S2. Design and train a machine learning model for automatically identifying and classifying main lobe signals and side lobe signals, perform adaptive main lobe position identification of the strongest signal in the initial noise source imaging, obtain the main lobe signal, and use statistical methods to verify the identification result in real time;

[0109] S3, in the initial noise source imaging, removing the contribution of the main lobe signal according to the real shape of the identified main lobe, and updating the remaining signal;

[0110] S4, repeating iterative steps S2-S3 until the remaining signal is lower than a preset threshold; wherein an adaptive control strategy is introduced in the iterative process to dynamically adjust the number of iterations, iteration step size and preset threshold according to signal changes;

[0111] S5. Superimpose all the removed main lobe signals and reconstruct the initial noise source imaging in combination with a multi-model projection technique to complete the removal of the side lobe effect.

[0112] Through the above steps, the CLEAN algorithm will be more robust and flexible, and can effectively remove the sidelobe effect, thereby improving the clarity and accuracy of noise source imaging. It intelligently handles the uncertainty and dynamics in the algorithm, and provides stronger support for accurate imaging in complex signal environments.

[0113] 5. High-resolution imaging

[0114] like Figure 3 As shown, based on the processed standard data, the inversion technology is used to reconstruct the spatial distribution image of the noise source, and then the adaptive algorithm is used to dynamically adjust the resolution of the reconstructed spatial distribution image to achieve high-resolution positioning of the noise source imaging.

[0115] 5.1 Image Reconstruction

[0116] A mathematical model of the noise source is established, which is usually a linear model, and the linear model is expanded into a nonlinear model according to the actual noise situation; the linear equation of the linear model is expressed as:

[0117] Ax=b

[0118] Where A is the system matrix, x is the noise source distribution to be solved, and b is the measured data;

[0119] Using weighted least squares combined with matrix decomposition techniques (such as QR decomposition or SVD to improve the stability and efficiency of the solution) and manifold learning techniques (such as manifold orthogonalization to optimize the performance in high-dimensional space), the weights are dynamically adjusted according to the signal quality or signal-to-noise ratio to calculate the linear equations and minimize the calculation error;

[0120] A regularization term (such as Tikhonov regularization) is introduced into the linear model, and an adaptive regularization strategy (such as automatic correlation maximization based on a Bayesian framework) for dynamically adjusting the regularization parameter is implemented according to the regularization term to deal with pathological problems or noise influences and complete the reconstruction of spatially distributed images. If nonlinear regularization, such as total variation regularization, is applied, image edges and textures can be better captured. The expression of the regularization term is:

[0121] min||Ax-b|| 2 +λLx|| 2

[0122] Among them, ||Ax-b|| 2 is the minimization error of the linear equation system, λ is the regularization parameter, L is the regularization matrix, and x is the noise source distribution to be solved;

[0123] The reconstruction quality can be improved by analyzing and comparing different reconstruction results using contrastive learning methods. The details of the reconstructed image can also be improved by using a generative adversarial network (GAN).

[0124] The above steps greatly improve the accuracy and efficiency of the noise source imaging inversion process, and can maintain high resolution and accuracy in complex and high-noise environments, while reducing the computational burden and improving adaptability.

[0125] 5.2 Adjust image resolution

[0126] Automatically select an initial resolution that meets the standard data characteristics using a machine learning algorithm; usually set to a medium resolution to balance computational burden and detail representation;

[0127] Using a convolutional neural network (which can better identify important details and textures in complex images) to identify detail areas in the reconstructed spatially distributed image, and combining signal characteristics and noise levels to determine areas in the detail areas that need to be enhanced;

[0128] Introducing local adaptive high-resolution reconstruction technology in the detailed area, using local reconstruction strategy to tailor resolution adjustment according to regional characteristics. At the same time, maintaining or reducing resolution in flat areas to reduce computational complexity and noise impact;

[0129] Adopting adaptive iterative algorithm combined with online learning mechanism (updating adaptive adjustment parameters in real time to improve dynamic response capability) to dynamically adjust the resolution and image reconstruction parameters of different areas;

[0130] A generative adversarial network is used to synthesize regions of different resolutions into a final image, and adaptive noise suppression and super-resolution techniques are integrated to post-process the final image (such as denoising and contrast enhancement to further improve image quality), eliminate artifacts caused by asynchronous splicing, improve visual quality, and complete high-resolution positioning of noise source imaging.

[0131] Therefore, the present invention provides a high-resolution positioning method for noise source imaging, which can achieve high-resolution positioning, improve the reliability and accuracy of noise detection, and make noise control and management more effective.

[0132] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0133] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A high-resolution positioning method for noise source imaging, characterized in that: The following steps are involved: Design sensor array layout, optimize array parameters, and select sensors; Use high-precision clocks to synchronize all sensors and perform synchronous acquisition, then record the signals of each sensor through multi-channel acquisition to obtain collected data; Performing signal processing on the collected data using adaptive filtering, machine learning calibration, time-frequency analysis, and intelligent beamforming technology to obtain standard data; Based on the standard data, high-resolution direction estimation is performed using an optimized multiple signal classification algorithm, and sidelobe effects are removed using an optimized CLEAN algorithm to complete the processing of the standard data; Based on the processed standard data, the spatial distribution image of the noise source is reconstructed using the inversion technology, and then the resolution of the reconstructed spatial distribution image is dynamically adjusted using an adaptive algorithm to achieve high-resolution positioning of the noise source imaging.

2. A high-resolution positioning method for noise source imaging according to claim 1, characterized in that: Design sensor array layout, optimize array parameters, and select sensors, including: Selecting an adaptive sensor array type according to the spatial dimension of the noise source location, and performing irregular distribution of sensor positions based on the selected sensor array type to obtain a sensor array; Increasing the array aperture of the sensor array and selecting and setting the sensor spacing to complete array parameter optimization; Based on the optimized sensor array, sensors with high sensitivity and wide bandwidth are selected, and the sensors are calibrated in amplitude and phase to achieve signal consistency and accuracy.

3. A high-resolution positioning method for noise source imaging according to claim 2, characterized in that: All sensors are synchronized using a high-precision clock and collected synchronously. The signals of each sensor are then recorded through multi-channel collection to obtain collected data, including: Select a sensor clock based on the GPS system, synchronize the sensor clock using the NTP protocol, and then calibrate the sensor clock regularly. Add a timestamp to the collected data of each sensor and broadcast a synchronization signal to complete the design of the synchronization mechanism and obtain the high-precision clock. Design a multi-channel data acquisition system for simultaneously recording multiple sensor signals, and introduce a machine learning technique for adaptively adjusting the sampling rate and data acquisition strategy in the multi-channel data acquisition system; Select high-speed data transmission protocols for real-time data transmission, and introduce edge computing technology in sensor nodes to reduce data transmission volume and improve response speed; Data redundancy and backup strategies as well as data verification mechanisms are introduced to perform real-time processing of transmitted data to obtain collected data.

4. A high-resolution positioning method for noise source imaging according to claim 3, characterized in that: The collected data is processed by using adaptive filtering, machine learning calibration, time-frequency analysis and intelligent beamforming technology to obtain standard data, including: An adaptive filter for dynamically adjusting the frequency response is selected, a multi-stage filtering strategy is designed in the adaptive filter and a nonlinear filtering technology for processing non-Gaussian noise is integrated to remove irrelevant frequency components; the multi-stage filtering strategy includes a broadband filter for removing external interference and a narrowband filter for refined processing; Designing an automatic calibration mechanism in which known frequency and amplitude are used as reference signals and a machine learning algorithm is used to predict and compensate for dynamically changing system errors to complete amplitude and phase calibration of sensor signals; Combine overlapping window weighting technology to extract the time-frequency characteristics of the signal, and use STFT algorithm or wavelet transform to analyze the time-frequency characteristics of the signal; The beam direction is dynamically adjusted using an adaptive beamforming algorithm, and a deep learning model for identifying the optimal direction and a multi-mode beam technology for enhancing signals in multiple directions are introduced into the adaptive beamforming algorithm to complete intelligent beamforming and obtain standard data.

5. A high-resolution positioning method for noise source imaging according to claim 4, characterized in that: High-resolution heading estimation using an optimized multiple signal classification algorithm, including: Integrate the standard data into a data matrix, perform block processing and covariance matrix calculation on the data matrix, and obtain a block covariance matrix; A randomized matrix decomposition method and a GPU algorithm are introduced to perform eigenvalue decomposition on the block covariance matrix to obtain a signal subspace and a noise subspace. The eigenvector corresponding to the maximum eigenvalue is selected through a dimensionality reduction technique to construct a signal subspace, and the remaining eigenvectors are used to construct a noise subspace. Construct a MUSIC spectrum function, scan possible noise source directions and calculate spectrum function values, and identify the peak position used to represent the noise source direction in the spectrum function values ​​in combination with a trained machine learning model; The Bayesian statistical method is used to analyze the peak position and estimate the direction of the noise source.

6. A high-resolution positioning method for noise source imaging according to claim 5, characterized in that: Use the optimized CLEAN algorithm to remove sidelobe effects, including: S1. Use the multi-band MUSIC algorithm to combine the information of different frequency bands to generate the initial noise source imaging; S2. Design and train a machine learning model for automatically identifying and classifying main lobe signals and side lobe signals, perform adaptive main lobe position identification of the strongest signal in the initial noise source imaging, obtain the main lobe signal, and use statistical methods to verify the identification result in real time; S3, in the initial noise source imaging, removing the contribution of the main lobe signal according to the real shape of the identified main lobe, and updating the remaining signal; S4, repeating iterative steps S2-S3 until the remaining signal is lower than a preset threshold; wherein an adaptive control strategy is introduced in the iterative process to dynamically adjust the number of iterations, iteration step size and preset threshold according to signal changes; S5. Superimpose all the removed main lobe signals and reconstruct the initial noise source imaging in combination with a multi-model projection technique to complete the removal of the side lobe effect.

7. A high-resolution positioning method for noise source imaging according to claim 6, characterized in that: The spatial distribution image of the noise source is reconstructed using inversion techniques, including: A linear model of the noise source is established, and the linear model is expanded into a nonlinear model according to the actual noise situation; the linear equation of the linear model is expressed as: Ax=b Where A is the system matrix, x is the noise source distribution to be solved, and b is the measured data; Using weighted least squares method combined with matrix decomposition technology and manifold learning technology, the weights are dynamically adjusted according to signal quality or signal-to-noise ratio to calculate the linear equations and minimize the calculation error; A regularization term is introduced into the linear model, and an adaptive regularization strategy for dynamically adjusting the regularization parameter is implemented according to the regularization term to deal with pathological problems or noise influences and complete the reconstruction of the spatial distribution image; the expression of the regularization term is: min||Ax-b|| 2 +λ||Lx|| 2 Among them, ||Ax-b|| 2 is the minimization error of the linear equation system, λ is the regularization parameter, L is the regularization matrix, and x is the noise source distribution to be solved; The contrastive learning method is used to improve the reconstruction quality by analyzing and comparing different reconstruction results.

8. A high-resolution positioning method for noise source imaging according to claim 7, characterized in that: Adaptive algorithms are used to dynamically adjust the resolution of the reconstructed spatial distribution image to achieve high-resolution positioning of noise source imaging, including: Automatically selecting an initial resolution that meets the standard data characteristics using a machine learning algorithm; Using a convolutional neural network to identify detail areas in the reconstructed spatial distribution image, and combining signal characteristics and noise levels to determine areas in the detail areas that need to be enhanced; Introducing a local adaptive high-resolution reconstruction technique in the detail area to customize the local resolution and maintain or reduce the resolution in the flat area; Adopting adaptive iterative algorithm combined with online learning mechanism to dynamically adjust the resolution and image reconstruction parameters of different areas; A generative adversarial network is used to synthesize regions of different resolutions into a final image, and adaptive noise suppression and super-resolution techniques are integrated to post-process the final image to eliminate artifacts caused by asynchronous stitching and complete high-resolution positioning of noise source imaging.

Images