Road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system

Abstract

The invention relates to the technical field of environmental noise monitoring, and discloses a road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system, which comprises an acoustic sensor module, a data collection and storage module, a data analysis and evaluation module, a three-dimensional sound field construction and display module and a traffic flow feature library. According to the system, a differential geometry principle is adopted, a sound field is regarded as a Riemannian manifold with a local microstructure, and accurate description of an irregular sound field is realized through a curvature self-adaptive sound field manifold construction technology; introducing a covariant derivative in Riemannian geometry, and constructing a sound propagation model adapted to a complex road environment; and realizing hierarchical decomposition and reconstruction of the sound field by using a multi-scale analysis theory. According to the method, the sound source positioning precision and the calculation efficiency are improved, seamless analysis from microcosmic to macroscopic is realized, an innovative solution is provided for traffic and noise collaborative management, and intelligent traffic and environmental noise management are effectively supported.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental noise monitoring, in particular to a road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system, which is applied to urban road traffic noise monitoring, evaluation and visual analysis. BACKGROUND

[0002] With the acceleration of urbanization and the rapid increase in the number of motor vehicles, road traffic noise has become one of the main environmental problems affecting the quality of urban residents' life. The traditional road traffic noise monitoring method mainly relies on single-point sound level meter measurement, which is difficult to fully reflect the noise distribution characteristics in complex road environment. The existing noise monitoring system usually has the following shortcomings: first, uniform grid points are used, which cannot be fine monitoring for areas with dramatic noise changes; second, the noise propagation model is too simplified, without considering the influence of temperature gradient, air flow and obstacles in complex road environment; third, it lacks comprehensive analysis ability of noise characteristics at different spatial scales, making it difficult to meet the noise management needs at vehicle level, road section level and regional level.

[0003] In the prior art, some systems try to use acoustic array technology for sound source positioning, but the accuracy is insufficient in complex multi-source environment; some systems use numerical simulation method to predict noise propagation, but the calculation efficiency is low and it is difficult to adapt to real-time changing traffic conditions. In addition, the existing three-dimensional sound field visualization technology is mostly limited to indoor acoustic field, and there is a lack of efficient sound field reconstruction method suitable for open road environment.

[0004] Therefore, it is urgent to develop a system that can accurately monitor, analyze and visualize road traffic noise, and realize intelligent processing of the whole process from noise data collection to three-dimensional sound field reconstruction. SUMMARY

[0005] The purpose of the present application is to provide a road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system, which aims to solve the problems of insufficient monitoring accuracy, simplified sound propagation model and poor scale adaptability in the prior art. The present application adopts the principle of differential geometry, regards the sound field as a Riemannian manifold with locally differentiable structure, and realizes the accurate reconstruction from discrete noise data to continuous three-dimensional sound field through three core technologies of curvature adaptive sound field manifold construction, covariant derivative sound propagation model and multi-scale sound field topological decomposition and reconstruction.

[0006] The present application proposes a big data driven dynamic pricing intelligent decision system, which includes:

[0007] An acoustic sensor module is configured to collect acoustic data and its location information in a road environment and send the acoustic data and its location information to a data collection and storage module; the data collection and storage module is communicatively connected to the acoustic sensor module and configured to receive the acoustic data and its location information and store the acoustic data and its location information in a local database; a data analysis and evaluation module is communicatively connected to the data collection and storage module and configured to perform time-frequency analysis on the acoustic data, construct a noise acoustic fingerprint map of a lane and a road network in a monitoring area according to noise acoustic fingerprint features and a sound field distribution of the acoustic sensor module; a three-dimensional sound field construction and display module is communicatively connected to the data collection and storage module and configured to construct a sound field manifold expression with adaptive curvature characteristics based on a Riemann manifold theory, generate a three-dimensional sound field with a locally differentiable structure, construct a sound propagation model adaptive to a complex road environment by introducing a covariant derivative in Riemann geometry, and realize hierarchical decomposition and reconstruction of the sound field by using a multi-scale analysis theory in differential geometry; and a traffic flow feature library is communicatively connected to the three-dimensional sound field construction and display module and configured to store traffic flow feature information, and the three-dimensional sound field construction and display module is configured to perform three-dimensional visualization on a traffic flow condition based on the traffic flow feature information.

[0008] Preferably, the acoustic sensor module includes a fixed sensor group installed on a traffic artery and a roadside to realize basic coverage of a monitoring area, a mobile sensor group installed on a motor vehicle to realize continuous monitoring of mobile vehicle noise, and a sensor communication unit configured to send acoustic data and location information collected by the fixed sensor group and the mobile sensor group to the data collection and storage module.

[0009] Preferably, the data collection and storage module includes a data receiving unit configured to receive acoustic data and location information sent by the acoustic sensor module in real time, a data preprocessing unit configured to perform denoising, calibration and time-frequency transformation on the acoustic data, and a hierarchical storage unit configured to store the processed data in different levels of storage systems according to timeliness, including a hot data storage area, a warm data storage area and a cold data storage area.

[0010] Preferably, the data analysis and evaluation module includes a time-frequency analysis unit configured to perform short-time Fourier transform and Mel frequency cepstral coefficient extraction on the acoustic data, an acoustic fingerprint feature extraction unit configured to extract an acoustic fingerprint feature vector from a time-frequency analysis result, a noise source separation unit configured to separate contributions of different sound sources, and an acoustic environment evaluation unit configured to calculate an acoustic environment evaluation index in a monitoring area in real time in combination with the acoustic fingerprint map and a preset noise tolerance value.

[0011] As preferred, the curvature adaptive sound field manifold construction implementation in the three-dimensional sound field construction and display module comprises: a local curvature calculation unit configured to construct a local neighborhood for each sampling point, calculate a gradient vector field and a Hessian matrix of the sound pressure value, and represent the local curvature characteristics of the sound field; an adaptive grid generation unit configured to design a grid density distribution function according to the local curvature characteristics, so that the grid density in a high curvature area is increased and the grid density in a low curvature area is reduced; a sound source fitting and positioning unit configured to fit the sound source position using a least square method based on the sound pressure distribution of the sampling points; and a manifold construction unit configured to divide the entire monitoring area into a plurality of overlapping local patches, construct a local parameterized expression in each patch, ensure the derivative continuity at the patch boundary through differential constraints, and form a complete sound field manifold expression.

[0012] As preferred, the covariant derivative sound propagation model implementation in the three-dimensional sound field construction and display module comprises: a Riemann metric construction unit configured to construct a sound wave propagation velocity tensor field based on environmental parameters, convert the velocity tensor field into a Riemann metric tensor, and represent the spatial anisotropy characteristics; a connection coefficient calculation unit configured to calculate Christoffel connection coefficients based on the Riemann metric tensor, and describe the influence of spatial curvature on the sound wave propagation direction; a covariant wave equation construction unit configured to replace the ordinary derivative in the traditional sound wave equation with a covariant derivative, introduce the metric tensor and the connection coefficients, and construct a modified sound wave equation; and a geodesic equation solving unit configured to construct a sound wave geodesic equation based on the covariant sound wave equation, solve the geodesic equation using a numerical integration method, and determine the distribution of sound energy at each point in space.

[0013] As preferred, the multi-scale sound field topological decomposition and reconstruction implementation in the three-dimensional sound field construction and display module comprises: a wavelet transform decomposition unit configured to perform multi-scale wavelet transform on the sound field manifold, and decompose the sound field into components of different frequencies and spatial scales; a feature operator construction unit configured to construct a sound field feature extractor based on a Laplace-Beltrami operator, and calculate eigenfunctions and eigenvalues on the sound field manifold; a topological feature extraction unit configured to analyze the topological features of the sound field at each scale, including critical points, watersheds and ridge lines, and construct a sound field feature atlas; and a multi-scale reconstruction unit configured to design a reconstruction algorithm based on a feature preservation principle, reconstruct a continuous sound field at an arbitrary scale through inverse wavelet transform, and ensure the preservation of key topological features in the reconstruction process.

[0014] As preferred, the traffic flow feature library comprises: a vehicle feature data area configured to store acoustic feature models of different vehicle models; a traffic state data area configured to store real-time traffic data such as vehicle flow, vehicle speed distribution and traffic state; and a historical mode data area configured to store historical traffic mode data of different time periods and different road sections, and support traffic prediction and sound field simulation.

[0015] As preferred, the system further comprises: edge computing nodes deployed at road sides and regional centers for data preprocessing and preliminary analysis; a central server cluster for performing complex computing tasks and large-scale data storage; and a communication network for connecting the acoustic sensor modules, the edge computing nodes and the central server cluster, supporting real-time transmission and processing of data.

[0016] As preferred, the system realizes traffic noise management and sound environment evaluation by: a real-time monitoring and early warning unit for monitoring real-time distribution of traffic noise, identifying vehicles exceeding the standard and noise hotspots, and issuing noise warnings; a traffic planning optimization unit for analyzing the relationship between traffic organization and noise, evaluating the noise impact of different signal timing schemes, and optimizing traffic flow organization; a noise prevention and control measure evaluation unit for simulating the effects of different noise prevention measures, optimizing the design and layout of soundproof walls, and evaluating the effects of road surface material replacement; and a city planning support unit for constructing a city acoustic environment map, evaluating the sound environment quality of different functional areas, and supporting sound environment zoning management.

[0017] The present application has the following beneficial effects:

[0018] 1. Accurate representation of complex sound field: By introducing the theory of Riemannian manifold, the sound field is expressed as a manifold with locally differentiable structure, realizing accurate description of irregularly shaped sound field. In complex scenes such as building reflection and terrain undulation, the prediction error is controlled within ±1.5dB, improving the accuracy by about 300% compared with traditional models.

[0019] 2. Adaptive computing efficiency: The curvature adaptive algorithm dynamically allocates computing resources according to the complexity of the sound field, optimizing computing efficiency while ensuring accuracy. Compared with the uniform grid method, the computing time is reduced by about 65%, while the sound field reconstruction accuracy is improved by about 40% in complex urban environments.

[0020] 3. Multi-scale comprehensive analysis capability: The multi-scale topological decomposition method realizes seamless analysis from micro to macro, enabling the system to simultaneously evaluate the impact of single vehicle noise and the cumulative effect of regional noise, providing a technical foundation for noise fine management.

[0021] 4. Noise and traffic correlation analysis: A bidirectional mapping relationship between noise characteristics and traffic state is established, which not only enables inference of traffic conditions from noise, but also predicts the impact of traffic changes on the sound environment, supporting intelligent traffic and environmental noise collaborative management. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 Fig. 1 is a schematic diagram of the overall architecture of the road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system of the present application;

[0023] Figure 2Structure diagram of the acoustic sensor module of the present application;

[0024] Figure 3 Structure diagram of the data collection and storage module of the present application;

[0025] Figure 4 Structure diagram of the data analysis and evaluation module of the present application;

[0026] Figure 5 Flow chart of the curvature adaptive sound field manifold construction unit in the three-dimensional sound field construction and display module of the present application;

[0027] Figure 6 Flow chart of the covariant derivative sound propagation model in the three-dimensional sound field construction and display module of the present application;

[0028] Figure 7 Flow chart of the multi-scale sound field topological decomposition and reconstruction in the three-dimensional sound field construction and display module of the present application;

[0029] Figure 8 Structure diagram of the traffic flow feature library of the present application;

[0030] Figure 9 Deployment architecture diagram of the edge computing node and the central server of the present application. DETAILED DESCRIPTION

[0031] Reference will now be made to the drawings, in which Figures 1-9 the specific embodiments of the present application will be described in detail.

[0032] As shown in Figure 1 , the road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system of the present application comprises an acoustic sensor module 10, a data collection and storage module 20, a data analysis and evaluation module 30, a three-dimensional sound field construction and display module 40, and a traffic flow feature library 50.

[0033] The acoustic sensor module 10 is used to collect acoustic data and its location information in the road environment, and send these data to the data collection and storage module 20. The data collection and storage module 20 is in communication connection with the acoustic sensor module 10, for receiving acoustic data and its location information, and storing these data to a local database. The data analysis and evaluation module 30 is in communication connection with the data collection and storage module 20, for carrying out time-frequency analysis processing on the acoustic data, and constructing a lane and road network noise acoustic fingerprint map of the monitoring area according to noise acoustic fingerprint features and acoustic field distribution of the acoustic sensor module. The three-dimensional acoustic field construction and display module 40 is in communication connection with the data collection and storage module 20, for constructing an acoustic field manifold expression with adaptive curvature characteristics based on the theory of Riemann manifold, generating a three-dimensional acoustic field with local differentiable structure, and constructing an acoustic propagation model adapting to complex road environment by introducing covariant derivatives in Riemann geometry, and realizing hierarchical decomposition and reconstruction of the acoustic field by using the multi-scale analysis theory in differential geometry. The traffic flow feature library 50 is in communication connection with the three-dimensional acoustic field construction and display module 40, for storing traffic flow feature information, and the three-dimensional acoustic field construction and display module 40 carries out three-dimensional visual display of the traffic flow condition based on the traffic flow feature information.

[0034] As shown in the figure, in an embodiment of the present application, the acoustic sensor module 10 includes a fixed sensor group 11, a mobile sensor group 12 and a sensor communication unit 13. Figure 2

[0035] The fixed sensor group 11 is installed on traffic arteries and road sides, for realizing basic coverage of the monitoring area. Preferably, the fixed sensor group 11 is arranged with a density of 8-12 sensors per kilometer of road, and is arranged more densely in intersection areas and uniformly distributed on straight road sections. The sensor installation height is usually 4-6 meters above ground, to avoid ground reflection and pedestrian interference. The fixed sensor is usually powered by mains power, and in some areas far from the mains, a solar auxiliary power system can be used.

[0036] The mobile sensor group 12 is installed on motor vehicles, for realizing continuous monitoring of mobile vehicle noise. Preferably, the mobile sensor group 12 is deployed on public vehicles such as buses and sanitation vehicles, covering 10%-20% of the vehicle fleet, to form a dynamic monitoring network. The mobile sensor is powered by vehicle-mounted power supply, and is equipped with a GPS positioning system to ensure accurate location labeling of the data.

[0037] The sensor communication unit 13 is used to send acoustic data and location information collected by the fixed sensor group 11 and the mobile sensor group 12 to the data collection and storage module 20. The fixed sensor group 11 preferentially uses wired network communication (such as industrial Ethernet), to ensure stability and high bandwidth of data transmission. While the mobile sensor group 12 uses 5G / 4G wireless communication technology, to support real-time transmission of data. ​

[0038] In a specific implementation, the acoustic sensor adopts a high-precision digital microphone with a sampling frequency of 48 kHz and a 24-bit quantization precision, and a dynamic range of more than 100 dB, which can cover the typical frequency range (20 Hz-20 kHz) of road traffic noise. The sensor is equipped with a wind shield and a waterproof shell to ensure normal operation under various weather conditions.

[0039] As shown in FIG. 1, in one embodiment of the present application, the data collection and storage module 20 includes a data receiving unit 21, a data preprocessing unit 22, and a hierarchical storage unit 23. Figure 3 The data receiving unit 21 is used to receive acoustic data and position information sent by the acoustic sensor module 10 in real time. This unit supports multi-protocol data access, including communication protocols such as MQTT and HTTP, and implements preliminary verification and integrity checking of data. Preferably, the data receiving unit 21 is configured with redundant receiving channels to ensure reliable data reception under network fluctuations.

[0040] The data preprocessing unit 22 is used for denoising, calibration, and time-frequency transformation of acoustic data. The denoising process uses an adaptive filtering algorithm to effectively suppress environmental background noise; the calibration process is based on reference microphone data from periodic calibration to correct sensor sensitivity; and the time-frequency transformation process uses the Short-Time Fourier Transform (STFT) technique to convert time-domain signals into time-frequency domain representations, facilitating subsequent analysis. The window size of the time-frequency transformation is usually set to 20-30 ms to balance the time and frequency resolution.

[0041] The hierarchical storage unit 23 is used to store the processed data into different levels of storage systems according to the timeliness, including hot data storage area, warm data storage area, and cold data storage area. The hot data storage area uses memory database technology to save complete data for the last 24 hours, supporting millisecond-level fast access; the warm data storage area uses a time series database (such as InfluxDB) to save data for the last 30 days, but uses appropriate downsampling strategies to reduce storage space; and the cold data storage area uses object storage technology to archive historical data for long-term storage, supporting on-demand access.

[0042] In addition, the hierarchical storage unit 23 also implements a multi-dimensional indexing mechanism, including time indexing, spatial indexing, and feature indexing, which greatly improves the data query efficiency. For spatial indexing, the Geohash encoding method is used to support fast retrieval based on location; and for feature indexing, the voiceprint feature vector is constructed to support similarity search.

[0043] As shown in FIG. 2, in one embodiment of the present application, the data analysis and processing module 30 includes a data retrieval unit 31, a feature extraction unit 32, and a data analysis unit 33.

[0044] Figure 4 ​As shown, in one embodiment of the present invention, the data analysis and evaluation module 30 includes a time-frequency analysis unit 31, a voiceprint feature extraction unit 32, a noise source separation unit 33, and a sound environment evaluation unit 34.

[0045] The time-frequency analysis unit 31 is used to perform short-time Fourier transform and Mel-frequency cepstral coefficient extraction on the acoustic data. The short-time Fourier transform uses a Hanning window with a window size of 1024 points and an overlap rate of 50%, providing good time-frequency resolution. Mel-frequency cepstral coefficient (MFCC) extraction simulates the auditory characteristics of the human ear, typically extracting the first 13 coefficients to effectively characterize the acoustic features of the noise. The time-frequency analysis results form a time-spectrum graph, visually displaying the time-varying spectral characteristics of the noise.

[0046] The voiceprint feature extraction unit 32 is used to extract voiceprint feature vectors from the time-frequency analysis results. This invention employs a multi-dimensional feature fusion method, comprehensively considering time-domain features (such as zero-crossing rate and short-time energy), frequency-domain features (such as spectral centroid and band energy ratio), and statistical features (such as kurtosis and skewness). The feature vector dimension is typically 64-128 dimensions, sufficient to characterize the feature differences between different types of traffic noise. For the feature extraction process, a sliding window technique is preferably used, with a window size of 100ms and a step size of 20ms, to ensure the continuity and stability of the features.

[0047] The voiceprint feature vector can be represented as:

[0048] ,

[0049] in: For voiceprint feature vectors, it is a 3D column vector; For the first Each feature component represents a different type of acoustic feature value; This is the dimension of the feature vector, typically ranging from 64 to 128, depending on the application requirements; This indicates the transpose of a vector, converting a row vector into a column vector.

[0050] Noise source separation unit 33 is used to separate the contributions of different sound sources. In complex road environments, noise typically originates from the superposition of multiple sound sources, such as engine noise, tire noise, horn noise, and wind noise. This invention employs blind source separation technology based on nonnegative matrix factorization (NMF), combined with a priori sound source models, to achieve effective separation of different noise sources. Mixed signals collected by one microphone (in (where the number of time sampling points is a given factor), sound source separation is achieved by solving the following optimization problem:

[0051] ,

[0052] in: For the observed signal matrix, For the number of microphones, This represents the number of time sampling points; This is a hybrid matrix representing the propagation characteristics from the sound source to the microphone; The source signal matrix contains Time series of individual sound sources; The estimated number of sound sources is typically taken as 3-5; The divergence function measures the difference between two matrices; in this embodiment, KL divergence is used. This is a regularization term used to constrain the sparsity or smoothness of the solution; and This is the regularization coefficient, which controls the weight of the regularization term and is typically set to 0.1-0.5. The optimization problem is solved iteratively using alternating least squares or multiplicative update rules.

[0053] The acoustic environment assessment unit 34 is used to combine the acoustic signature map with preset noise tolerance values ​​to calculate the acoustic environment assessment index within the monitoring area in real time. This invention constructs a comprehensive acoustic environment assessment system that considers not only the traditional equivalent sound level (…). In addition to the existing indicators, the acoustic environment assessment index also incorporates time-varying characteristic indicators (such as volatility), spectral characteristic indicators (such as low-frequency proportion), and subjective perception indicators (such as annoyance level). The formula for calculating (SoundEnvironmentIndex) is:

[0054] ,

[0055] in: The sound environment evaluation index is dimensionless and ranges from 0 to 1. The lower the value, the better the sound environment quality. Sound level is a sound level indicator that reflects the absolute intensity of noise. It is a volatility index that reflects the time-varying characteristics of noise. It is a spectral indicator that reflects the frequency distribution characteristics of noise; This is a perceived indicator that reflects the subjective level of annoyance caused by noise. , , and For the corresponding weight coefficients, satisfying In typical applications, it can be set to , , , .

[0056] Sound level index The calculation is based on a comparison between the equivalent sound level and the allowable value:

[0057] ,

[0058] in: Equivalent sound level, measured in dB(A), represents the average noise energy over a certain period of time; and These are the minimum and maximum permissible sound levels, both measured in dB(A), and are typically determined according to national standards, such as those for residential areas at night. dB(A), dB(A). When hour, ;when hour, .

[0059] like Figures 5 to 7 As shown, the three-dimensional sound field construction and display module 40 is the core innovative part of this invention, which realizes accurate reconstruction from discrete noise data to a continuous three-dimensional sound field based on differential geometry theory. This module includes a curvature adaptive sound field manifold construction unit 41, a covariant derivative sound propagation model unit 42, and a multi-scale sound field topology decomposition and reconstruction unit 43.

[0060] like Figure 5 As shown, the curvature adaptive sound field manifold building unit 41 includes a local curvature calculation unit 411, an adaptive mesh generation unit 412, a sound source fitting and localization unit 413, and a manifold building unit 414.

[0061] The local curvature calculation unit 411 is used to construct a local neighborhood for each sampling point, calculate the gradient vector field and Hessian matrix of the sound pressure value, and characterize the local curvature characteristics of the sound field. First, for each sampling point... Determine its local neighborhood Typically, a spherical region with radius r is selected (r is usually set to 5-10 meters). Then, the sound pressure function is fitted within the local neighborhood. Calculate its gradient vector field and the Hesse matrix :

[0062] ,

[0063] in: For position sound pressure function The gradient vector field is a three-dimensional vector; is the sound pressure function, representing the sound pressure value at various points in space, with the unit being dB(A). It is a spatial position vector. , , These are the coordinates in a Cartesian coordinate system, in meters. , , Representing the sound pressure function right , , The partial derivatives reflect the rate of change of sound pressure in each direction; This represents the transpose of a vector.

[0064] ,

[0065] in: For position sound pressure function The Hessian matrix is ​​a A real symmetric matrix; elements of the matrix , , This represents the second derivative of the sound pressure function along each coordinate axis, reflecting the magnitude of the curvature; The mixed partial derivative terms represent the coupling changes between different directions. The eigenvalues ​​and eigenvectors of the Hessian matrix represent the principal curvature and principal direction, respectively, and are used to guide adaptive mesh generation.

[0066] The adaptive mesh generation element 412 is used to design the mesh density distribution function based on local curvature characteristics, increasing the mesh density in high-curvature regions and decreasing the mesh density in low-curvature regions. Mesh density function. Defined as:

[0067] ,

[0068] in: For position The grid density at a given location is expressed in points per square meter. and These are the minimum and maximum mesh densities, typically set to [value]. Points per square meter (corresponding to a grid size of approximately 3 meters). Points per square meter (corresponding to a grid size of approximately 0.5 meters); For position The curvature index at a point is usually taken as the maximum absolute value of the principal curvature, reflecting the degree of drastic change in the sound field at that point; The mapping function maps the curvature to the [0,1] interval and is used to control the distribution of mesh density.

[0069] Mapping function Use the sigmoid function:

[0070] ,

[0071] in: The parameter used to control the curve steepness is usually set to 5. The larger the value, the more sensitive the grid density is to changes in curvature. This is the curvature threshold, typically set to 0.3, indicating that the mesh density begins to increase significantly when the curvature exceeds this value; The base of the natural logarithm is used. Based on the grid density function, an adaptive grid partitioning is achieved using a quadtree or octree structure, ensuring that computational resources are concentrated in regions where the sound field changes drastically.

[0072] The sound source fitting and localization unit 413 is used to fit the sound source location based on the sound pressure distribution at the sampling points using the least squares method. For the point sound source model, the sound pressure attenuation with distance can be expressed as:

[0073] ,

[0074] in: For position The sound pressure level at a certain location is measured in dB(A). For reference distance The sound pressure level at each location is expressed in dB(A). The location of the sound source is a three-dimensional coordinate vector. For position To the sound source Euclidean distance, in meters; For reference distance, it is usually taken as 1 meter; The air absorption coefficient is measured in dB / meter and typically ranges from 0.005 to 0.02, depending on frequency, temperature, and humidity. Represent the logarithmic function to the base 10. Given a set of sampling points. The location of the sound source is determined by minimizing the following objective function. and sound source intensity :

[0075] ,

[0076] in: For position The measured sound pressure level at the location is expressed in dB(A). For the first The location coordinates of each sampling point; The total number of sampling points is represented by ; the meanings of the other symbols are the same as in the previous equation. This optimization problem is nonlinear and is typically solved using the Levenberg-Marquardt algorithm or Newton's method. For multi-source problems, it is preferable to use the RANSAC (Random Sample Consensus) algorithm combined with cluster analysis, first grouping the sampling points and then fitting the sound sources within each group.

[0077] Manifold building unit 414 is used to divide the entire monitoring area into multiple overlapping local patches. A local parametric representation is constructed within each patch, and derivative continuity at patch boundaries is ensured through differential constraints, forming a complete acoustic field manifold representation. Each local patch is represented by a parametric surface.

[0078] ,

[0079] in: For parameter space The sound pressure level is expressed in dB(A). The coordinates are in a two-dimensional parameter space, and are usually normalized to interval; For a control point, it is a three-dimensional vector containing its position coordinates and sound pressure value; and For B-spline basis functions, and The order of the B-spline is given, preferably cubic B-splines. ,supply Continuity; and For the index of the control point, , . This represents a summation operation, accumulating the products of all control points and their corresponding basis functions. The number of control points is adaptively set based on the complexity of the local region, typically 1. to .

[0080] The complete sound field manifold can be represented as a set of all local patches:

[0081] ,

[0082] in: For a complete sound field manifold; For the first The parameterized representation of a local patch is a function; For the first The parameter space domain of a patch is typically a two-dimensional planar region. For the mapping function from parameter space to physical space, the two-dimensional parameters are... Mapped to three-dimensional physical space; For the index of patches; The total number of patches depends on the size and complexity of the monitored area. Connections between adjacent patches use a shared control point method to ensure... Continuity, meaning that the sound pressure gradient is continuous at the patch boundary.

[0083] like Figure 6 As shown, the covariant derivative sound propagation model unit 42 includes a Riemannian metric construction unit 421, a connection coefficient calculation unit 422, a covariant sound wave equation construction unit 423, and a geodesic equation solving unit 424.

[0084] The Riemannian metric building unit 421 is used to construct a sound wave propagation velocity tensor field based on environmental parameters, converting the velocity tensor field into a Riemannian metric tensor to characterize spatial anisotropy. In a non-uniform flow medium, the sound wave propagation velocity can be expressed in tensor form:

[0085] ,

[0086] in: For position The sound wave propagation speed tensor at a point is a The matrix is ​​expressed in meters per second; For position The base sound velocity at a given location (affected by factors such as temperature and humidity), measured in meters per second; for identity matrix; For position The velocity vector at a given location is a three-dimensional vector, measured in meters per second. This tensor describes the speed at which sound waves propagate in different directions.

[0087] Base speed of sound It can be calculated using the following formula:

[0088] ,

[0089] in: The basic speed of sound, measured in meters per second; Temperature is measured in degrees Celsius; 331.3 is... The speed of sound at that time is the reference value, in meters per second; 0.606 is the temperature coefficient, in meters per second. Flow velocity vector Data can be obtained through environmental monitoring equipment or predicted through computational fluid dynamics models.

[0090] Based on the sound wave propagation speed tensor, the Riemannian metric tensor It can be represented as:

[0091] ,

[0092] in: For position The components of the Riemannian metric tensor at that location. , representing the tensor's first Line number Column elements; For position The equivalent sound speed scalar at that point is usually taken as ; For the Kronecker delta function, when The value is 1 if the condition is met, and 0 otherwise. and These are the velocity vectors. The and the One component; For the magnitude of the flow velocity, i.e. The Riemannian metric tensor describes the ease with which sound waves propagate in all directions, providing a foundation for subsequent calculations of covariant derivatives.

[0093] The connection coefficient calculation unit 422 is used to calculate the Christopher connection coefficient based on the Riemannian metric tensor, describing the effect of spatial curvature on the direction of sound wave propagation. Christopher connection coefficient. The calculation formula is:

[0094] ,

[0095] in: For Christopher's contact coefficient, This describes the rate of change of the vector field during parallel transmission; Riemannian metric tensor The reverse, satisfying , in For Kronecker delta function; Represents the metric tensor components coordinates The partial derivatives; This represents the i-th coordinate component, corresponding to the Cartesian coordinate system. , , Summation convention: In the above formula, the index that appears repeatedly... This indicates that the index is summed, i.e. The connection coefficient is typically calculated using numerical methods, such as finite difference or spline interpolation, to ensure accurate calculation at discrete sampling points.

[0096] The covariant acoustic wave equation building unit 423 is used to replace the ordinary derivative in the traditional acoustic wave equation with a covariant derivative, introduce a metric tensor and connection coefficients, and construct a modified acoustic wave equation. The traditional acoustic wave equation is:

[0097] ,

[0098] in: Sound pressure level, measured in Pascals (Pa). Speed ​​of sound, measured in meters per second; Time, in seconds; Let be the second partial derivative of sound pressure with respect to time, representing the change in acceleration of sound pressure; The Laplace operator is defined as follows: , represents the spatial second derivative of sound pressure.

[0099] In curvature space, the Laplace operator Laplace-Beltrami operator Alternative:

[0100] ,

[0101] in: The Laplace-Beltrami operator acting on the sound pressure function The result; Riemannian metric tensor The determinant, i.e. ; It is the square root of the determinant and the volume element scaling factor in curvature space; Represents coordinates The partial derivatives; The inverse of the Riemannian metric tensor; Represents the sound pressure function coordinates Partial derivatives. Summation convention: recurring indices. and This indicates summing these indicators, i.e. .

[0102] The corrected equation for sound waves is:

[0103] ,

[0104] This equation takes into account the effect of spatial curvature on sound wave propagation, and more accurately describes the sound propagation phenomenon in complex road environments. The symbols have the same meaning as before.

[0105] The geodesic equation solving unit 424 is used to construct the sound wave geodesic equation based on the covariant sound wave equation. The numerical integration method is used to solve the geodesic equation to determine the distribution of sound energy at various points in space. The geodesic equation describes the propagation path of sound waves in Riemannian space.

[0106] ,

[0107] in: The position coordinates of the first One portion, , respectively corresponding , , For geodesic parameters, representing the distance along the geodesic line; coordinates For parameters The second derivative of represents the acceleration of the geodesic; The Christopher contact coefficient, as defined above; and Coordinates and For parameters The first derivative of represents the velocity component of the geodesic. Summation convention: recurring indices and This indicates summing these indicators, i.e. The geodesic equations are solved using the fourth-order Runge-Kutta method, with a step size typically set to 0.1-0.5 meters, which can be adaptively adjusted according to the curvature.

[0108] Sound energy at position The calculations at that location took into account factors such as geometric diffusion, air absorption, and reflection / diffraction.

[0109] ,

[0110] in: For position The sound energy at a given location, measured in joules per square meter; The energy of the sound source is expressed in joules. To reach the location The number of geodesic lines; For the first The reflection / diffraction coefficients of the path are dimensionless and range from [value missing]. 0 represents complete reflection, and 0 represents complete absorption; This is the air absorption coefficient, expressed in units of 1 / m. For the first The length of a geodesic line, in meters; This indicates the energy loss caused by air absorption; This indicates energy decay caused by geometric diffusion; It is the solid angle of a sphere. This represents a summation operation, which accumulates the energy contributed by all geodesics.

[0111] Sound pressure level can be obtained from energy conversion:

[0112] ,

[0113] in: For position The sound pressure level at the location, measured in dB; For position The sound energy at a given location, measured in joules per square meter; Reference energy, corresponding to reference sound pressure. Pa, the calculation formula is: ,in air density, Speed ​​of sound; This represents the logarithmic function with base 10.

[0114] like Figure 7 As shown, the multi-scale sound field topology decomposition and reconstruction unit 43 includes a wavelet transform decomposition unit 431, a feature operator construction unit 432, a topology feature extraction unit 433, and a multi-scale reconstruction unit 434.

[0115] Wavelet transform decomposition unit 431 is used to perform multi-scale wavelet transform on the sound field manifold, decomposing the sound field into components of different frequencies and spatial scales. For a manifold defined on... sound field function on Its wavelet transform can be expressed as:

[0116] ,

[0117] in: For function In scale and location Wavelet transform coefficients at; The wavelet basis functions are preferably Meyer wavelets or second-generation wavelets; This is a scale parameter that controls the spatial scale of the analysis; its unit is the same as that of the spatial coordinates. The position parameter represents the wavelet center position; Points on the manifold and points The geodetic distance between them is in the same unit as the spatial coordinates; Indicated throughout the entire manifold Integrals on; The area element is on the manifold. Wavelet decomposition is typically performed at 5-7 scale levels, covering spatial scales from 1 meter (vehicle near field) to 500 meters (area noise).

[0118] Feature operator construction unit 432 is used to construct a sound field feature extractor based on the Laplace-Beltramian operator, and to calculate the eigenfunctions and eigenvalues ​​on the sound field manifold. Laplace-Beltramian operator The characteristic equation is:

[0119] ,

[0120] in: This refers to the Laplace-Beltrami operator defined earlier; Let be the i-th eigenfunction, defined on the manifold superior; These are the corresponding eigenvalues, usually sorted by size. The eigenfunctions form an orthogonal basis that satisfies ,in This is the Kronecker delta function.

[0121] Eigenfunctions can be used for spectral decomposition of sound fields:

[0122] ,

[0123] in: Let be the sound field function defined on the manifold; For function Eigenfunctions The projection on is calculated using the following formula: ; For the i-th eigenfunction at point The value at; This represents the summation of the contributions of all eigenfunctions. In practical calculations, the first 100-500 eigenfunctions are usually sufficient to characterize the main features of the sound field.

[0124] The topological feature extraction unit 433 is used to analyze the topological features of the sound field at various scales, including critical points, watersheds, and ridges, to construct a sound field feature map. Critical points include local maxima (noise sources), local minima (quiet zones), and saddle points (bifurcation points of propagation paths), which are determined by solving for the zeros of the gradient field.

[0125] ,

[0126] in: For function At point The gradient vector at a given point represents the rate of change of the function in each direction; Let be a point on the manifold. The point is considered to be zero when all components of the gradient vector are zero. These are critical points. The type of a critical point is determined by the eigenvalues ​​of the Hessian matrix: all negative values ​​indicate a maximum point, all positive values ​​indicate a minimum point, and a combination of positive and negative values ​​indicates a saddle point.

[0127] A watershed is the boundary line of the sound field gradient flow, defined as a curve extending along the gradient direction from a saddle point. Ridges are lines connecting local maximum sound pressure levels, representing the main propagation path of noise. These topological features constitute a sound field feature map, describing the structural characteristics of the sound field.

[0128] The multi-scale reconstruction unit 434 is used to design a reconstruction algorithm based on the feature preservation principle. It reconstructs the continuous sound field at arbitrary scales through inverse wavelet transform, ensuring the preservation of key topological features during the reconstruction process. The reconstruction algorithm first determines the target scale. Then select the set of topological features to retain. The inverse wavelet transform can be expressed as:

[0129] ,

[0130] in: For target scale The reconstructed sound field function is as follows; These are the wavelet transform coefficients defined earlier; For position ,scale wavelet basis functions at point The value at; This indicates summing over all scales greater than or equal to the target scale; This represents the summation over all points on the manifold. In practical calculations, the integral is usually converted into a discrete summation, where both the point location and the scale parameter take a finite number of discrete values.

[0131] To ensure the preservation of topological features, constraints are introduced during the reconstruction process:

[0132] ,

[0133] in: Representation function The set of topological features, including feature points, feature lines, etc.; The set of key topological features to be retained is determined by application requirements. This constraint is achieved through iterative optimization, adjusting the reconstruction function at each step until all key topological features are preserved. Preferably, the L-BFGS (Limited-memory Broyden–Fletcher–Goldfarb–Shanno) algorithm is used for optimization, with convergence conditions being a reconstruction error of less than 2 dB or 100 iterations.

[0134] In practical applications, the reconstruction scale can be flexibly selected according to different analysis needs: micro scale (1-10 meters) is suitable for vehicle noise identification and excessive monitoring; meso scale (10-100 meters) is suitable for road noise assessment and short-term prediction; macro scale (100-500 meters) is suitable for regional noise distribution and environmental impact assessment; and urban scale (>500 meters) is suitable for urban planning and policy making.

[0135] like Figure 8As shown, in one embodiment of the present invention, the traffic flow feature database 50 includes a vehicle feature data area 51, a traffic status data area 52, and a historical pattern data area 53.

[0136] The vehicle feature data area 51 is used to store acoustic feature models for different vehicle types. This data area contains acoustic feature templates for various types of vehicles (such as small passenger cars, medium-sized passenger cars, large passenger cars, small trucks, medium-sized trucks, large trucks, motorcycles, etc.), including the spectral characteristics and intensity characteristics of components such as engine noise, tire noise, and aerodynamic noise. Preferably, the feature model for each vehicle type includes acoustic characteristics under four operating conditions: idling, acceleration, constant speed, and deceleration, as well as noise characteristics at different vehicle speeds (such as 30km / h, 50km / h, 70km / h, etc.). These feature models are constructed using a large amount of real-vehicle test data and are regularly updated to adapt to changes in vehicle technology.

[0137] Traffic status data area 52 is used to store real-time traffic data such as traffic flow, vehicle speed distribution, and traffic conditions. This data area is integrated with traffic monitoring systems (such as loop detectors, video detectors, and radar detectors) to receive and store traffic flow parameters in real time, including flow rate (vehicles / hour), vehicle speed (km / h), density (vehicles / km), and occupancy rate (%). In addition, it includes traffic status classifications (such as smooth flow, light congestion, moderate congestion, and severe congestion) and vehicle type composition ratios (such as the percentage of large vehicles). This data is typically statistically analyzed and stored with a time granularity of 5 minutes or 15 minutes.

[0138] Historical pattern data area 53 stores historical traffic pattern data for different time periods and road sections, supporting traffic prediction and sound field simulation. This data area extracts typical temporal patterns (such as morning peak, evening peak, and off-peak) and spatial patterns (such as main roads, secondary roads, and local roads) through statistical analysis of long-term accumulated traffic data. These pattern data include weekday patterns, weekend patterns, holiday patterns, and traffic patterns under special weather conditions. Based on these historical patterns and combined with real-time traffic data, short-term (15 minutes to 2 hours) and medium-term (2 to 24 hours) traffic condition changes can be predicted, providing input for noise prediction.

[0139] Preferably, the traffic flow feature database 50 adopts a distributed database architecture, supporting high-concurrency access and fast querying. Data storage uses a spatiotemporal index structure, facilitating retrieval by time range and spatial region. The data update strategy combines real-time updates (for traffic status data) and periodic updates (for vehicle feature data and historical pattern data) to ensure data timeliness and accuracy.

[0140] like Figure 9As shown, in one embodiment of the present invention, the system further includes an edge computing node 61, a central server cluster 62, and a communication network 63.

[0141] Edge computing nodes 61 are deployed on roadsides and in regional centers for data preprocessing and preliminary analysis. Roadside edge nodes are typically deployed in existing facilities such as traffic signal control cabinets and monitoring poles, with each node covering multiple sensors within a radius of 300-500 meters. The hardware configuration of roadside nodes is typically an 8-core CPU, 16GB of RAM, and 128GB of storage, sufficient to support data preprocessing and preliminary analysis tasks. Regional edge nodes are deployed in locations such as traffic management sub-centers and environmental monitoring stations, covering multiple roadside nodes within a radius of 3-5 kilometers. Regional nodes have higher configurations, typically with a 16-core CPU, 64GB of RAM, 2TB of storage, and equipped with GPU accelerator cards, capable of performing moderately complex analysis tasks.

[0142] The central server cluster of 62 servers is used for complex computing tasks and large-scale data storage. The compute servers are configured with 64 CPU cores, 256GB of memory, and multiple GPUs / TPUs for acceleration, handling computationally intensive tasks such as sound field modeling and multi-scale analysis. The storage servers employ a distributed storage array, providing petabyte-level storage capacity to support long-term storage of historical data and big data analysis. The application servers are configured with 32 CPU cores and 128GB of memory, responsible for web services, API interfaces, user authentication, and business logic processing. The server cluster employs load balancing and high availability design to ensure stable system operation.

[0143] The communication network 63 connects the acoustic sensor module 10, edge computing nodes 61, and the central server cluster 62, supporting real-time data transmission and processing. The network architecture employs a multi-layered design: wired connections (such as industrial Ethernet) are prioritized between sensors and roadside edge nodes to ensure data transmission stability; 5G / 4G wireless connections are used between mobile sensors and edge nodes; fiber optic leased lines or 5G networks are used between roadside nodes and regional nodes; and high-speed fiber optic networks are used between regional nodes and the central server. At the communication protocol level, the MQTT protocol is used to transmit real-time data, the HTTP / HTTPS protocol to transmit query and control commands, and the WebSocket protocol to support bidirectional real-time communication.

[0144] The system of this invention realizes traffic noise management and acoustic environment assessment through the following units: real-time monitoring and early warning unit, traffic planning optimization unit, noise prevention and control measures assessment unit, and urban planning support unit.

[0145] The real-time monitoring and early warning unit is used to monitor the real-time distribution of traffic noise, identify vehicles exceeding noise standards and noise hotspots, and issue noise warnings. This unit analyzes voiceprint characteristics to identify abnormal noise events in real time, such as vehicles exceeding noise standards (e.g., motor vehicles with modified exhaust pipes), construction noise, and noise from sudden events. Noise exceeding standards is determined based on preset thresholds, such as 70 dB(A) during the day and 55 dB(A) at night in residential areas; exceeding the threshold for more than 10 seconds is considered exceeding the standard. Warning information is displayed in real-time through the system interface and can be pushed to the mobile terminals of relevant management departments, supporting timely intervention and handling.

[0146] The traffic planning optimization unit analyzes the relationship between traffic organization and noise, assesses the noise impact of different signal timing schemes, and optimizes traffic flow organization. This unit constructs a traffic noise impact model, quantifying the influence of different traffic parameters (such as flow rate, speed, and vehicle type composition) on noise levels. For example, analysis shows that on urban roads, increasing the average vehicle speed from 30 km / h to 50 km / h can reduce the equivalent noise level by approximately 2-3 dB; while for every 10% increase in the proportion of large vehicles, the noise level increases by approximately 1.5 dB. Based on these relationships, traffic signal timing can be optimized to reduce acceleration and deceleration processes and lower noise impact; vehicle routes can be adjusted to guide large vehicles to road sections with less impact on noise-sensitive areas.

[0147] The noise control measure evaluation unit is used to simulate the effects of different noise reduction measures, optimize the design and layout of sound barriers, and evaluate the effects of road material replacement. Based on a three-dimensional sound field model, this unit can simulate the effects of various noise control measures, such as sound barriers (of different heights, materials, and locations), green belts, and low-noise road surfaces. By comparing changes in noise distribution before and after the implementation of measures, the effectiveness of the control measures is quantified, supporting cost-benefit analysis. For example, the evaluation shows that a 3-meter-high vertical sound barrier can reduce noise by approximately 8-10 dB at a distance of 20 meters from the road; while porous asphalt pavement can reduce tire-road noise by approximately 3-5 dB compared to ordinary asphalt pavement. These quantitative results provide a scientific basis for noise control decisions.

[0148] The urban planning support unit is used to construct urban acoustic environment maps, assess the acoustic environment quality of different functional zones, and support acoustic environment zoning management. This unit combines noise monitoring data with GIS (Geographic Information System) data to build urban acoustic environment maps, visually displaying noise distribution. The acoustic environment map includes not only current status assessments but also planning predictions, enabling the assessment of the potential impact of new roads and urban development projects on the acoustic environment. Furthermore, based on the acoustic environment quality assessment, it supports the division and management of acoustic environment functional zones, such as determining noise control targets and management measures for different functional zones like special quiet zones (around schools and hospitals), residential areas, commercial areas, and industrial areas.

[0149] Through the collaborative work of the aforementioned units, the system of this invention achieves intelligent management of the entire process from noise monitoring to management decision-making, providing an innovative solution for urban sound environment governance.

[0150] This invention is not limited to the above embodiments. Those skilled in the art can make various equivalent transformations and modifications to this invention without departing from the spirit and scope of this invention, and such equivalent transformations and modifications should be included within the protection scope of the claims of this invention.

Claims

1. A road traffic noise intelligent monitoring and three-dimensional sound field reconstruction system, characterized in that, include: An acoustic sensor module is used to collect acoustic data and location information in the road environment, and send the acoustic data and location information to a data collection and storage module; The data collection and storage module is communicatively connected to the acoustic sensor module, and is used to receive the acoustic data and its location information, and store the acoustic data and its location information in a local database; The data analysis and evaluation module is communicatively connected to the data collection and storage module and is used to perform time-frequency analysis processing on the acoustic data and construct lane and road network noise soundprint maps of the monitoring area based on noise soundprint characteristics and sound field distribution of the acoustic sensor module. A three-dimensional sound field construction and display module, communicatively connected to the data collection and storage module, is used to construct a sound field manifold expression with adaptive curvature characteristics based on Riemannian manifold theory, generate a three-dimensional sound field with locally differentiable structure, and construct a sound propagation model adapted to complex road environments by introducing covariant derivatives in Riemannian geometry, and realize hierarchical decomposition and reconstruction of the sound field using multi-scale analysis theory in differential geometry; and a traffic flow feature library, communicatively connected to the three-dimensional sound field construction and display module, is used to store traffic flow feature information, and the three-dimensional sound field construction and display module performs three-dimensional visualization of traffic flow conditions based on the traffic flow feature information.

2. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The acoustic sensor module includes: Fixed sensor arrays are installed along major traffic arteries and roadsides to provide basic coverage of the monitoring area; A mobile sensor group, installed on a motor vehicle, is used to continuously monitor the noise of the moving vehicle; and a sensor communication unit is used to send the acoustic data and location information collected by the fixed sensor group and the mobile sensor group to the data collection and storage module.

3. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The data collection and storage module includes: The data receiving unit is used to receive acoustic data and location information sent by the acoustic sensor module in real time; The data preprocessing unit is used to perform noise reduction, calibration and time-frequency transformation processing on the acoustic data; and the hierarchical storage unit is used to store the processed data in different levels of storage system according to timeliness, including hot data storage area, warm data storage area and cold data storage area.

4. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The data analysis and evaluation module includes: The time-frequency analysis unit is used to perform short-time Fourier transform and Mel frequency cepstral coefficient extraction on the acoustic data; The voiceprint feature extraction unit is used to extract voiceprint feature vectors from time-frequency analysis results; The noise source separation unit is used to separate the contributions of different sound sources; and the sound environment evaluation unit is used to combine the sound pattern map with the preset noise tolerance value to calculate the sound environment evaluation index in the monitoring area in real time.

5. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The curvature-adaptive sound field manifold construction method in the three-dimensional sound field construction and display module includes: The local curvature calculation unit is used to construct a local neighborhood for each sampling point, calculate the gradient vector field and Hess matrix of the sound pressure value, and characterize the local curvature characteristics of the sound field. An adaptive mesh generation element is used to design a mesh density distribution function based on local curvature characteristics, thereby increasing the mesh density in high curvature regions and decreasing the mesh density in low curvature regions. The sound source fitting and localization unit is used to fit the sound source location based on the sound pressure distribution of the sampling points using the least squares method; and the manifold construction unit is used to divide the entire monitoring area into multiple overlapping local patches, construct a local parameterized expression within each patch, and ensure the continuity of derivatives at the patch boundaries through differential constraints to form a complete sound field manifold expression.

6. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The implementation methods of the covariant derivative sound propagation model in the three-dimensional sound field construction and display module include: Riemannian metric building unit is used to construct a sound wave propagation velocity tensor field based on environmental parameters, and to convert the velocity tensor field into a Riemannian metric tensor to characterize spatial anisotropy. The connection coefficient calculation unit is used to calculate the Christopher connection coefficient based on the Riemannian metric tensor, which describes the effect of spatial curvature on the direction of sound wave propagation. The covariant acoustic wave equation construction unit is used to replace the ordinary derivative in the traditional acoustic wave equation with a covariant derivative, introduce the metric tensor and connection coefficient, and construct a modified acoustic wave equation; and the geodesic equation solving unit is used to construct the acoustic wave geodesic equation based on the covariant acoustic wave equation, solve the geodesic equation using the numerical integration method, and determine the distribution of sound energy at various points in space.

7. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The implementation methods for multi-scale sound field topology decomposition and reconstruction in the three-dimensional sound field construction and display module include: The wavelet transform decomposition unit is used to perform multi-scale wavelet transform on the sound field manifold, decomposing the sound field into components of different frequencies and spatial scales. Feature operator construction unit, used to construct a sound field feature extractor based on the Laplace-Beltramian operator, and to calculate the eigenfunctions and eigenvalues ​​on the sound field manifold; The topology feature extraction unit is used to analyze the topology features of the sound field at various scales, including critical points, watersheds and ridges, and to construct a sound field feature map; and the multi-scale reconstruction unit is used to design a reconstruction algorithm based on the feature preservation principle, and to reconstruct the continuous sound field at any scale through inverse wavelet transform, ensuring the preservation of key topology features during the reconstruction process.

8. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The traffic flow feature database includes: The vehicle feature data area is used to store acoustic feature models of different vehicle models; The traffic status data area is used to store real-time traffic data such as traffic flow, vehicle speed distribution, and traffic status; while the historical pattern data area is used to store historical traffic pattern data for different time periods and different road sections, supporting traffic prediction and sound field simulation.

9. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The system also includes: Edge computing nodes, deployed on the roadside and in regional centers, are used for data preprocessing and preliminary analysis; A central server cluster is used to perform complex computing tasks and large-scale data storage; and a communication network is used to connect the acoustic sensor module, the edge computing node and the central server cluster, supporting real-time data transmission and processing.

10. The intelligent road traffic noise monitoring and three-dimensional sound field reconstruction system according to claim 1, characterized in that, The system achieves traffic noise management and acoustic environment assessment in the following ways: The real-time monitoring and early warning unit is used to monitor the real-time distribution of traffic noise, identify vehicles exceeding the standard and noise hotspots, and issue noise warnings. The traffic planning optimization unit is used to analyze the relationship between traffic organization and noise, evaluate the noise impact of different signal timing schemes, and optimize traffic flow organization. The noise control measures evaluation unit is used to simulate the effects of different noise control measures, optimize the design and layout of sound barriers, and evaluate the effects of road surface material replacement. And an urban planning support unit, used to construct urban acoustic environment maps, assess the acoustic environment quality of different functional areas, and support acoustic environment zoning management.

Images