Low-altitude aircraft environment noise modeling method and system

Abstract

The invention relates to the technical field of noise modeling, in particular to a low-altitude aircraft environment noise modeling method and system. Sound source directivity data of the low-altitude aircraft at different positions under the spherical coordinate system are reprojected to a hemispherical surface and then converted into sound source directivity tensors; converting the environmental topographic information and the building geometric information into topographic layout tensors; fusing the advanced features of the sound source directivity tensor, the advanced features of the terrain layout tensor, the frequency features of the low-altitude aircraft at the current moment and the flight height features of the low-altitude aircraft at the current moment to obtain multi-modal potential features at the current moment; and inputting the multi-modal potential features at the current moment into the diffusion model, and outputting a noise intensity graph of the low-altitude aircraft at the current moment. According to the method, complex multi-modal conditions such as noise source characteristics, terrain layout and flight height of the low-altitude aircraft can be accurately captured and fused, and the efficiency, precision and generalization ability of noise modeling are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of noise modeling technology, and in particular to a method and system for modeling environmental noise of low-altitude aircraft. Background Technology

[0002] With the rapid development of aviation technology, the application of low-altitude aircraft is becoming increasingly widespread. However, the environmental noise generated during their low-altitude flight not only disturbs residents' lives and the ecological environment but may also affect the acoustic environment of sensitive areas such as airport perimeters. Therefore, noise issues are receiving increasing attention. Environmental noise modeling for low-altitude aircraft is a core technology for addressing this problem, accurately simulating and predicting noise intensity maps of aircraft at different locations and in different environments.

[0003] Traditional methods for modeling environmental noise in low-altitude aircraft primarily rely on hardware measurements such as array microphones and computer numerical simulations. Array microphone measurement is a traditional noise measurement method that measures the noise level of an aircraft as it flies by by deploying microphone arrays at designated locations. This method requires a large amount of hardware, and the equipment needs to be reconfigured for each measurement, making it costly and inflexible. For example, when measuring noise in an urban environment, multiple microphone arrays need to be deployed at different locations to capture the noise generated by the aircraft at different positions. However, this method can only measure in a limited number of locations and cannot comprehensively capture the noise distribution of the aircraft throughout its flight path. Moreover, because the placement of microphone arrays is limited by geographical conditions and environmental factors, it is difficult to conduct large-scale noise measurements in complex urban environments.

[0004] Computer numerical simulation methods simulate sound wave propagation and calculate the sound pressure level at any observation point. Common numerical simulation methods include wave-like methods and geometric methods. Wave-like methods, such as the finite element method, boundary element method, and finite difference time-domain method, can accurately simulate the propagation process of sound waves, but these methods require solving a large number of equations, resulting in extremely high computational costs, especially when simulating large-scale urban environments, where they consume significant time and computational resources. Geometric methods, such as the Gaussian beam tracing method, calculate noise distribution by simulating the geometric propagation path of sound waves. Although more efficient than wave-like methods, they still suffer from high computational complexity and cost when dealing with complex geometric environments and dynamic scenes.

[0005] To address the limitations of traditional methods, such as high hardware measurement costs, poor flexibility, and high computational complexity in numerical simulation, hindering large-scale deployment, the development of deep learning technology has brought new opportunities for environmental noise modeling of low-altitude aircraft. Some studies have pioneered the use of the U-Net architecture, taking a fixed sound source model and a 2D map of urban buildings as input to output ground noise distribution. In the field of communication propagation, similar approaches also focus on 2D ground signal mapping, and the sound source morphology is limited to simple models such as monopoles and dipoles.

[0006] While deep learning technology has brought new opportunities, existing deep learning methods still have certain limitations when dealing with multimodal conditions (such as different aircraft types, flight altitudes, terrain layouts, and atmospheric conditions), making it difficult to accurately capture the coupling effects of complex acoustic features and environmental factors. Firstly, existing deep learning methods can only process preset sound sources and cannot generalize to arbitrarily complex aircraft. This is because the sound source data used in the training of existing deep learning models are often preset sound sources of specific types with fixed parameters, whose noise characteristics are relatively simple and fixed. However, actual low-altitude aircraft vary greatly in engine type, number and speed of propellers, and flight attitude due to different aircraft types (such as helicopters, fixed-wing UAVs, and multi-rotor UAVs). Therefore, when faced with new or complex aircraft not present in the training data, it is impossible to accurately capture their noise characteristics, thus hindering generalization. Secondly, existing methods mostly use 2D urban building maps as input. This two-dimensional information can only reflect the planar location and general outline of buildings. However, noise is affected by the three-dimensional occlusion of buildings during propagation. Different facades of buildings reflect noise, and facades with different angles and materials have different reflection coefficients and reflection directions, resulting in complex three-dimensional characteristics in the distribution of reflected noise in space. Since the model does not consider these three-dimensional factors and only calculates based on two-dimensional information, it cannot accurately simulate the changes in noise intensity caused by occlusion and reflection in three-dimensional space. Therefore, the generated 2D ground map cannot truly reflect the actual noise distribution. In addition, the network structure design of U-Net focuses more on learning the spatial mapping of input data and lacks a dedicated mechanism to model and integrate the nonlinear coupling relationships between these multiple factors. Therefore, U-Net cannot fully capture the coupling relationship between noise and multiple factors such as terrain and atmosphere, and cannot fully explore the complex correlation between the undulation of terrain, the dynamic changes of the atmosphere, and the noise propagation characteristics, resulting in poor accuracy of the generated noise distribution map. Summary of the Invention

[0007] Therefore, the technical problem to be solved by the present invention is to overcome the shortcomings of existing deep learning-based low-altitude aircraft environmental noise modeling methods, such as single sound source processing, neglect of three-dimensional influencing factors, and U-Net's inability to fully capture the coupling relationship between noise and multiple factors such as terrain and atmosphere, resulting in poor accuracy of low-altitude aircraft environmental noise modeling.

[0008] To address the aforementioned technical problems, this invention provides a method for modeling environmental noise of low-altitude aircraft, comprising: Construct a spherical coordinate system for the environment under test, and convert the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system into a sound source directivity tensor; The three-dimensional environmental terrain information and three-dimensional building geometry information of the environment under test are converted into a terrain layout tensor. High-level features of the sound source directivity tensor and terrain layout tensor are extracted respectively, as well as the frequency and flight altitude features of the low-altitude aircraft at the current moment. By fusing the high-level features of the sound source directivity tensor and the terrain layout tensor, the frequency features and flight altitude features of the low-altitude aircraft at the current moment, the multimodal latent features at the current moment are obtained. Based on the multimodal potential features at the current moment, obtain the noise intensity map of the low-altitude aircraft at the current moment.

[0009] Preferably, the process of obtaining the multimodal latent features at the current time includes: After concatenating the sound source directivity tensor and the terrain layout tensor, the output features of each encoder layer of the first U-Net model are obtained through the encoder of the first U-Net model. The frequency and altitude of the low-altitude aircraft at the current moment are extracted by passing them through their corresponding learnable embedding layers. The output features of the last encoder layer of the first U-Net model, the frequency features of the low-altitude aircraft at the current moment, and the flight altitude features are fused by element-level addition and then passed through the bottleneck layer of the first U-Net model to obtain the bottleneck layer features of the first U-Net model. The bottleneck layer features of the first U-Net model and the output features of each encoder layer of the first U-Net model are passed through the decoder of the first U-Net model to obtain the multimodal latent features at the current time.

[0010] Preferably, obtaining the noise intensity map of the low-altitude aircraft at the current moment based on the multimodal latent features at the current moment includes: Using the current multimodal latent features as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment; where... This represents the preset total number of time steps.

[0011] Preferably, the step of using the current multimodal latent features as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment, including: A temporal embedding vector for each time step is generated using sinusoidal position encoding; The first Noise intensity map at time step 1, the first time step 2 The temporal embedding vector at the nth time step and the multimodal latent features at the current time step are obtained through the nth time step. The reverse diffusion process at the nth time step yields the nth time step. The noise intensity map at each time step includes: A temporal embedding vector for each time step is generated using sinusoidal position encoding; The first Noise intensity map at time step 1, the first time step 2 The temporal embedding vector at the nth time step and the multimodal latent features at the current time step are obtained through the nth time step. The reverse diffusion process at the nth time step yields the nth time step. The noise intensity map at each time step includes: The first Noise intensity map at time step 1, the first time step 2 The temporal embedding vectors at each time step and the multimodal latent features at the current time step are fused using element-wise addition to obtain the first... The fused feature maps at each time step; among them... , For time step index, the first The noise intensity map at each time step was obtained by sampling from a standard normal distribution; The first The fused feature map at the nth time step is passed through the encoder of the second U-Net model to obtain the nth time step. The second U-Net model at the first time step The target feature map output by each encoder layer; where... This represents the total number of encoder layers in the second U-Net model. The first The second U-Net model at the first time step The target feature map output by the encoder layer is passed through a multi-head self-attention block to obtain the first... The second U-Net model at the first time step Multi-head self-attention feature maps corresponding to each encoder layer; The first The second U-Net model at the first time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the bottleneck layer of the second U-Net model to obtain the first encoder layer. Bottleneck feature map of the second U-Net model at each time step; The first Bottleneck feature map of the second U-Net model at each time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the decoder of the second U-Net model with residual connections to obtain the first encoder layer. Noise intensity map at each time step; Use the noise intensity map at time step 0 as the noise intensity map of the low-altitude aircraft at the current time.

[0012] Preferably, the step of placing the first Bottleneck feature map of the second U-Net model at each time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the decoder of the second U-Net model with residual connections to obtain the first encoder layer. The noise intensity map at each time step includes: The first The bottleneck feature map of the second U-Net model after upsampling at the first time step and the first time step The second U-Net model at the 1st time step The multi-head self-attention feature maps corresponding to the encoder layers are concatenated along the channel dimension and then passed through the first decoder layer of the second U-Net model to obtain the first... Decoding fusion feature map of the first decoder layer of the second U-Net model at each time step; The first The decoding fusion feature map of the first decoder layer of the second U-Net model at the 1st time step, The bottleneck feature map after upsampling the second U-Net model at the nth time step is used to obtain the nth time step through element-wise addition. The target fusion feature map of the first decoder layer of the second U-Net model at each time step; The first The second U-Net model at the 1st time step The target fusion feature map upsampled by the decoder layer and the first decoder layer The second U-Net model at the 1st time step The output of the encoder layer is the first... After the target feature maps at each time step are concatenated along the channel dimension, they are processed by the second U-Net model. The decoder layer is obtained to get the first... The second U-Net model at the 1st time step Decoding fusion feature maps of each decoder layer; where... , For decoder layer index; The first The second U-Net model at the 1st time step Decoding fusion feature map of the decoder layer, the first The second U-Net model at the 1st time step The target fusion feature map upsampled from the decoder layer is used to obtain the th element-wise addition. The second U-Net model at the 1st time step Target fusion feature map of each decoder layer; The first The second U-Net model at the 1st time step The target fusion feature map of the decoder layer is used as the first decoder layer. Noise intensity map at each time step.

[0013] Preferably, an aircraft noise assessment model is constructed based on a multimodal conditional encoder and a diffusion model; The loss function of the aircraft noise assessment model includes: The diffusion denoising loss between the noise intensity map of the low-altitude aircraft at the current moment and the actual noise intensity map; The smoothed L1 loss between the multimodal latent features and the true noise intensity map at the current time.

[0014] Preferably, the sound source directivity data includes: the frequency characteristics and intensity characteristics of the sound.

[0015] Preferably, a spherical coordinate system is constructed for the environment to be tested. The sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system are reprojected onto a hemisphere, and the hemisphere data is converted into a sound source directivity tensor.

[0016] Preferably, noise intensity maps of low-altitude aircraft at various times are obtained. For each noise intensity map belonging to the same flight altitude, noise footprint processing is performed to obtain multiple 2D images belonging to the same flight altitude. Multiple 2D images at the same flight altitude are interpolated to generate a 3D noise distribution visualization.

[0017] The present invention also provides a low-altitude aircraft environmental noise modeling system, comprising: The first quantization module is used to construct a spherical coordinate system of the environment under test, and convert the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system into a sound source directivity tensor. The second quantization module is used to convert the three-dimensional environmental terrain information and three-dimensional building geometric information of the environment under test into a terrain layout tensor. The feature extraction module is used to extract high-level features of the sound source directivity tensor and terrain layout tensor, as well as the frequency and flight altitude features of the low-altitude aircraft at the current moment. The fusion module is used to fuse the high-level features of the sound source directivity tensor and the terrain layout tensor, the frequency features and flight altitude features of the low-altitude aircraft at the current moment, to obtain the multimodal latent features at the current moment. The prediction module is used to obtain the noise intensity map of the low-altitude aircraft at the current moment based on the multimodal potential features at the current moment.

[0018] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: This invention discloses a method and system for modeling environmental noise of low-altitude aircraft. The invention constructs a spherical coordinate system for the environment under test, reprojects the sound source directivity data of the low-altitude aircraft at different locations within this coordinate system onto a hemisphere, and converts it into a two-dimensional tensor. This encompasses the complex noise characteristics arising from different aircraft types and flight attitudes, achieving generalization for any complex aircraft. Simultaneously, environmental terrain information and building geometry information are converted into a terrain layout tensor, fully incorporating information such as terrain undulations and the three-dimensional structure of buildings. This accurately reflects the three-dimensional occlusion and reflection effects of different facades during noise propagation, avoiding biases caused by two-dimensional information input. Through the encoder of the first U-Net model of the aircraft noise assessment model, high-level geometric features are extracted from the concatenation of the sound source directivity tensor and the terrain layout tensor, combined with the current-moment frequency characteristics extracted through a learnable embedding layer. The characteristics of flight altitude, after element-level additive fusion, are passed through a bottleneck layer and then input into the decoder of the first U-Net model along with the output characteristics of the encoder layer to obtain multimodal latent features. Finally, these features are input into the diffusion model to output a noise intensity map. This method can deeply explore the complex relationship between environmental factors such as terrain and atmosphere and noise propagation characteristics, accurately model the nonlinear coupling relationship between multiple factors, and input the multimodal latent features at the current moment into the diffusion model of the aircraft noise assessment model to output the noise intensity map of the low-altitude aircraft at the current moment. The diffusion model conforms to the physical logic of noise spreading outward from the sound source and being gradually modulated by environmental factors. Through the backward denoising process, it gradually learns the mapping from noise to the real distribution, optimizes the calculation process of noise estimation, significantly improves the efficiency of noise modeling, and can effectively improve the prediction accuracy of the noise intensity map of low-altitude aircraft. Attached Figure Description

[0019] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart illustrating a method for modeling environmental noise of low-altitude aircraft according to the present invention.

[0020] Figure 2 This is a schematic diagram of the modeling and visualization results of drone noise tracking in urban Shanghai. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0022] Reference Figure 1 As shown, this embodiment provides a method for modeling environmental noise of low-altitude aircraft, including: To address the shortcomings of existing technologies in handling multimodal conditions, this invention designs a unique multimodal sensing module and a multimodal condition encoding mechanism. This mechanism can accurately capture and fuse complex multimodal conditions such as the noise source characteristics of low-altitude aircraft, terrain layout, and flight altitude, effectively improving the accuracy and generalization ability of noise modeling and better meeting the complex needs of urban low-altitude aircraft noise management. The multimodal sensing module is shown in steps S1-S2.

[0023] Step S1: Construct a spherical coordinate system for the environment under test. Reproject the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system onto a hemisphere and convert it into a sound source directivity tensor. After acquiring the sound source directivity data of the low-altitude aircraft at different spatial locations using a spherical coordinate system, a reprojection technique is employed to map it onto a hemisphere, achieving dimensionality reduction of the aircraft's sound source spatial characteristics. This aims to achieve dimensionality matching with input data from another modality (a 128*128 city model). This method innovatively transforms high-dimensional spatial characteristics into a multi-layer image format (128*128*2) and inputs it into a deep learning model for subsequent processing.

[0024] In this embodiment, specifically, the sound source directivity data includes: the frequency characteristics and intensity characteristics of the sound (such as sound pressure level and sound power).

[0025] This embodiment extracts the acoustic source characteristics of low-altitude aircraft based on their structural parameters (including key components such as rotors and fuselage) and operating conditions (such as flight speed and load conditions), and characterizes them from two dimensions: acoustic source directivity and frequency characteristics. Based on the original sound source model data in spherical coordinates, it is mapped to a hemisphere using spatial reprojection technology; further, a projection algorithm is used to transform the hemispherical data into a 128×128 dimensional tensor. This is used to quantitatively describe the directional distribution pattern of sound sources in space.

[0026] Frequency characteristic characterization then directly models the frequency parameters of the sound source in scalar form, using them as independent dimensional inputs to reflect the spectral characteristics of the sound source.

[0027] Traditional technologies have significant limitations in describing the sound source characteristics of low-altitude aircraft: they only abstract the sound source information into isolated scalars (e.g., "flying in the positive x direction at a speed of 5 m / s" is represented by only two scalars, "5 m / s" and "+1"). This simplification completely loses the details of the radiation directionality of the sound source in three-dimensional space and cannot reflect the differences in noise intensity in various directions under different aircraft types and flight attitudes (e.g., pitch angle, sideslip). (For example, the noise radiation intensity of a propeller aircraft in the forward direction may differ by several times from that in the rearward direction.)

[0028] This invention constructs a spherical coordinate system for the environment under test, reprojecting the sound source directivity data of a low-altitude aircraft at any location onto a hemisphere to form a three-dimensional universal directivity spherical model. This model quantifies the radiation intensity of the sound source in each direction in three-dimensional space using a sound source directivity tensor. It does not rely on preset parameters specific to any aircraft type and can be adapted to any aircraft (fixed-wing, multi-rotor, helicopter, etc.), truly achieving a universal and refined characterization of sound source properties.

[0029] This invention chooses to input frequency characteristics in scalar form, primarily based on considerations of both the dynamic characteristics of aircraft noise and practical engineering applications. Although the frequency characteristics of an aircraft change dynamically during flight, this change can be fully reflected through discretization in the time dimension. When the aircraft moves along a predetermined trajectory, the model makes independent predictions based on the frequency characteristics of the current position at each moment. The input frequency scalar strictly corresponds to the noise characteristics at that moment, and the frequency parameters at adjacent moments do not conflict because they belong to different time nodes. For the broadband characteristics of aircraft noise, the model adopts a single-input, single-frequency processing method. This design conforms to the common practice in industry and academia of breaking down broadband characteristics into 1-20 discrete frequencies for analysis, while also providing good flexibility. To fully evaluate the noise characteristics over a wide frequency range, a corresponding number of model instances can be run simultaneously. Each instance focuses on processing one discrete frequency component, and the prediction results at each frequency are then integrated to obtain the full-band analysis conclusion.

[0030] This approach ensures accurate capture of dynamic frequency changes, maintains the integrity of broadband characteristic analysis, and is consistent with conventional methods in engineering practice, demonstrating the organic unity of theoretical rigor and practical feasibility.

[0031] The sound generated by low-altitude aircraft has three dimensions: time, space, and frequency. Among them, the directivity of the sound source can be represented by a directional sphere to show the propagation characteristics of spherical waves in space. The time characteristic is implicit in the coordinates of the aircraft, which affects the sound propagation distance and radiation range. As the aircraft's coordinates move, the surrounding urban environment moves accordingly (the aircraft is always located directly above the city center). The frequency characteristic is separated as a dimension because different frequencies of sound have different attenuation factors in the atmosphere and different absorption factors by walls (i.e., different reflection coefficients). These multimodal inputs are all closely related to acoustics.

[0032] Step S2: Convert the 3D environmental terrain information and 3D building geometry information of the environment to be measured into a terrain layout tensor; In this embodiment, the digital elevation model of a geographic information system is used to obtain three-dimensional environmental terrain information; The distribution, shape, and height features of buildings are extracted using public datasets such as OpenStreetMap, Google Maps, and Open3Dhk to obtain three-dimensional geometric information of buildings. The two types of information mentioned above are combined using modeling software to form a geometric city model in formats such as .stl and .vtk. This model is then divided into equal-sized areas along the horizontal direction to form a standardized model. Finally, it is converted into a 128×128 horizontal 2D terrain layout tensor through quantization. ; At the same time, the terrain height is projected onto the ground plane to form a structured representation, which facilitates processing by the aircraft noise assessment model.

[0033] Step S3: Extract high-level features from the sound source directivity tensor and terrain layout tensor, as well as the frequency and altitude features of the low-altitude aircraft at the current moment. The quantized terrain data is encoded, and key geometric features are extracted to form a compact representation for noise propagation modeling. These features include information such as the distribution and height of buildings and the undulation of the terrain, which can reflect the impact of terrain on sound wave propagation.

[0034] Step S4: Fuse the high-level features of the sound source directivity tensor and the terrain layout tensor, the frequency features of the low-altitude aircraft at the current moment, and the flight altitude features to obtain the multimodal latent features at the current moment; Step S5: Based on the multimodal potential features at the current moment, obtain the noise intensity map of the low-altitude aircraft at the current moment.

[0035] In this embodiment, preferably, the process of obtaining the multimodal latent features at the current time includes: like Figure 1 As shown, Figure 1 The multimodal conditional encoder is the encoder of the first U-Net model, and the multimodal conditional decoder is the decoder of the first U-Net model. The sound source directivity tensor Topographic layout tensor After splicing, the output features of each encoder layer of the first U-Net model are obtained through the encoder of the first U-Net model; The frequency and altitude of the low-altitude aircraft at the current moment are extracted by passing them through their corresponding learnable embedding layers. The output features of the last encoder layer of the first U-Net model, the frequency features of the low-altitude aircraft at the current moment, and the flight altitude features are fused element-wise by addition. This fusion is then passed through the bottleneck layer of the first U-Net model to obtain the bottleneck layer features of the first U-Net model, as shown in the formula: , in, Features of the bottleneck layer of the first U-Net model. The encoder for the first U-Net model. This is element-level addition. This is the learnable embedding layer corresponding to the frequency of the low-altitude aircraft at the current moment. The frequency of the low-altitude aircraft at the current moment. This is the learnable embedding layer corresponding to the current flight altitude of the low-altitude aircraft. This represents the current flight altitude of the low-altitude aircraft. For splicing operations, Represents the directivity tensor of the sound source. Represents the terrain layout tensor.

[0036] The bottleneck layer features of the first U-Net model and the output features of each encoder layer of the first U-Net model are passed through the decoder of the first U-Net model to obtain the multimodal latent features at the current time.

[0037] In this embodiment, preferably, obtaining the noise intensity map of the low-altitude aircraft at the current moment based on the multimodal latent features at the current moment includes: Using the current multimodal latent features as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment; where... This represents the preset total number of time steps.

[0038] The diffusion model adopts an encoder-decoder structure based on the U-Net model, which includes an encoder. and decoder with residual connection The diffusion model uses a predefined sequence. The noise intensity map is gradually added to the step size for forward diffusion; during backward diffusion, a neural network is trained to gradually denoise the noise to restore the original noise intensity map.

[0039] In this embodiment, specifically, the step of inputting the multimodal latent features at the current moment into the diffusion model and outputting the noise intensity map of the low-altitude aircraft at the current moment includes: Using the current multimodal latent features as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment; where... This represents the preset total number of time steps.

[0040] In this embodiment, preferably, the current multimodal latent features are used as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment, including: like Figure 2 As shown, Figure 2 The noise estimation encoder is the encoder of the second U-Net model, and the noise estimation decoder is the decoder of the second U-Net model with residual connections. A temporal embedding vector for each time step is generated using sinusoidal position encoding; The first Noise intensity map at each time step , No. The time embedding vector at each time step Multimodal latent features at the current moment Through the first The reverse diffusion process at the nth time step yields the nth time step. The noise intensity map at each time step includes: The first Noise intensity map at each time step , No. The time embedding vector at each time step Current multimodal latent features By fusing elements through element-level addition, we obtain the first... The fused feature maps at each time step; among them... , For time step index, the first The noise intensity map at each time step was obtained by sampling from a standard normal distribution; The first The fused feature map at the nth time step is passed through the encoder of the second U-Net model to obtain the nth time step. The second U-Net model at the first time step The target feature map output by each encoder layer; where... The total number of encoder layers in the second U-Net model is given by the formula: The formula is: , in, For the first The second U-Net model at the first time step The target feature map output by each encoder layer For noise estimation encoders, For the first Noise intensity map at each time step, For the current time-mode multimodal latent features, For the first Temporal embedding vectors at each time step.

[0041] The first The second U-Net model at the first time step The target feature map output by the encoder layer is enhanced by multi-head self-attention blocks to model spatial dependencies, associating each spatial location with conditionally relevant regions to better learn acoustic interference textures, thus obtaining the first... The second U-Net model at the first time step The multi-head self-attention feature map corresponding to each encoder layer is given by the formula: , in, For bullish self-attention blocks, For the first The second U-Net model at the first time step The multi-head self-attention feature map corresponding to each encoder layer For the query weight matrix, This is the key weight matrix. Value is the weight matrix.

[0042] The first The second U-Net model at the first time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the bottleneck layer of the second U-Net model to obtain the first encoder layer. Bottleneck feature map of the second U-Net model at each time step; The first Bottleneck feature map of the second U-Net model at each time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the decoder of the second U-Net model with residual connections to obtain the first encoder layer. The noise intensity map at each time step is given by the following formula: , in, For the first Noise intensity map at each time step, The denoising process for the decoder of the second U-Net model with residual connections. For the first The second U-Net model at the first time step Multi-head self-attention feature maps corresponding to each encoder layer.

[0043] The noise intensity map at time step 0 This is a noise intensity diagram of a low-altitude aircraft at the current moment.

[0044] In this embodiment, preferably, the loss function of the aircraft noise assessment model includes: The diffusion denoising loss between the noise intensity map of the low-altitude aircraft at the current moment and the actual noise intensity map; the diffusion denoising loss is used to optimize the model's ability to predict the noise components added in each forward diffusion step.

[0045] The smoothed L1 loss between the multimodal latent features and the true noise intensity map at the current time is expressed by the following formula: , in, The multimodal conditional alignment loss is the smoothing L1 loss between the multimodal latent features and the true noise intensity map at the current time. The height of the actual noise intensity map at the current moment. The width of the current noise intensity map. The height index of the pixel. The width index of the pixel. The first of the multimodal latent features at the current time Line 1 Column pixel values, The first element in the current time noise intensity graph Line 1 Column pixel values, This is a smoothed L1 loss function.

[0046] Multimodal conditional alignment loss is used to ensure that the encoded conditions accurately capture physically meaningful patterns in the input modalities, in order to force... and Structural consistency.

[0047] In this embodiment, the training process of the aircraft noise assessment model includes: Initialize the parameters of the diffusion model, including the weights of the second U-Net model encoder and the decoder with residual connections, the parameters of the learnable embedding layers, etc.

[0048] During the forward diffusion process, for the input noise intensity map, a predefined sequence is used. Step size, gradually add Gaussian noise to obtain Noise intensity map corresponding to each time step with added Gaussian noise ; Based on the sound source directivity tensor, terrain layout tensor, frequency and flight altitude of the low-altitude aircraft at the current moment output by the multimodal perception module, the multimodal potential features at the current moment are obtained. from Noise intensity map corresponding to each time step with added Gaussian noise Begin, through The backdiffusion process is performed step by step to gradually remove noise and generate the final noise intensity map. ; The diffusion denoising loss and multimodal conditional alignment loss are calculated. The diffusion denoising loss measures the difference between the noise components predicted by the model and the actual noise components; the multimodal conditional alignment loss measures the predicted multimodal latent features at the current time step. Compared with the actual noise intensity map Structural consistency between them; Based on the calculated loss value, update the parameters of the diffusion model, including the weights of the U-Net encoder and decoder, the parameters of the learnable embedding layers, etc. Repeat the forward diffusion, conditional encoding, backward diffusion, loss calculation, and parameter update steps until the aircraft noise assessment model converges, i.e., the loss value reaches the preset threshold or the training reaches the preset number of iterations. The trained aircraft noise assessment model is then used as the target aircraft noise assessment model.

[0049] In this embodiment, preferably, noise intensity maps of low-altitude aircraft at various times are obtained, and for each noise intensity map belonging to the same flight altitude, noise footprint processing is performed to obtain multiple 2D images belonging to the same flight altitude. Noise footprint refers to the effective range of noise generated by a low-altitude aircraft at a certain altitude and time on the ground (or observation plane). The core function of noise footprint mapping is to transform the numerical values ​​in the noise intensity map into the spatial noise influence range. That is, by setting a threshold, the outline and distribution of the effective noise area are extracted from the original noise intensity map to generate a 2D image that focuses only on the noise influence range.

[0050] Multiple 2D images at the same flight altitude are interpolated to generate a 3D noise distribution visualization.

[0051] In this embodiment, the noise footprint reconstruction module is used to generate detailed 2D and 3D images of noise distribution visualization, providing intuitive decision-making basis for urban planning and noise mitigation strategies. The 2D visualization displays noise levels at different height levels, where layer H=0 represents ground noise, while higher layers display noise levels on building surfaces and in the air. Noise data at different height levels are visualized separately to obtain 2D noise distribution images. Layer H=0 represents ground noise, and layers H=50, 100, 150, and 200 represent noise levels at different heights.

[0052] 3D visualization generates a spatial representation of noise distribution in 3D by interpolating hierarchical noise data (2D noise data) onto a 3D environmental model. 3D visualization provides a more intuitive view of the propagation and distribution of noise in space. Analyzing the visualization results of noise distribution allows for the identification of noise hotspots—areas with high noise levels. These hotspots are crucial for evaluating the effectiveness of noise mitigation strategies and developing targeted noise control measures.

[0053] This invention aims to address several key issues in the field of environmental noise modeling for low-altitude aircraft. First, addressing the drawbacks of high cost and poor flexibility of traditional hardware measurement methods, this invention proposes a new method for modeling environmental noise in low-altitude aircraft, aiming to effectively reduce measurement costs and improve modeling flexibility. Second, addressing the high computational complexity of numerical simulation methods, this invention innovatively introduces a diffusion model, optimizing the noise estimation calculation process and significantly improving modeling efficiency, making it more suitable for noise simulation in large-scale urban environments.

[0054] This invention optimizes the noise estimation computation process through an innovative multimodal sensing module and diffusion model, significantly improving the efficiency of noise modeling. Compared with traditional numerical simulation methods, this invention can complete large-scale urban environment noise simulation in a shorter time, providing a more efficient solution for urban low-altitude aircraft noise management. This invention effectively reduces the cost of noise modeling. Traditional hardware measurement methods require numerous microphone arrays and complex equipment configurations, resulting in high costs. This invention utilizes the powerful computational capabilities of deep learning models and diffusion models, reducing dependence on hardware and lowering measurement costs. Simultaneously, this method also improves the model's flexibility and scalability, enabling it to better adapt to different application scenarios and needs.

[0055] The multimodal perception module designed in this invention can accurately capture various complex multimodal conditions such as the noise source characteristics, terrain layout, and flight altitude of low-altitude aircraft. By effectively fusing and encoding these conditions, the model can more accurately simulate the noise propagation process, improving the accuracy and generalization ability of noise modeling. This allows the invention to better adapt to the complex and ever-changing needs of urban low-altitude aircraft noise management, providing more reliable data support for formulating effective noise control strategies.

[0056] The noise footprint reconstruction module proposed in this invention can generate detailed 2D and 3D images of noise distribution visualization, providing intuitive decision-making basis for urban planning and noise mitigation strategies. By analyzing the visualization results of noise distribution, noise hotspots can be clearly identified, the effectiveness of noise mitigation strategies can be evaluated, and targeted noise control measures can be formulated. This is of great significance to city managers and policymakers, and helps to achieve refined management of low-altitude aircraft noise in cities.

[0057] In this embodiment, noise hotspots are defined as areas with noise levels above 70 dB. The proposed low-altitude aircraft environmental noise modeling method is used to model and track UAV noise in Shanghai's urban area. Figure 2 As shown, Figure 2 A schematic diagram illustrating the modeling and visualization results of drone noise tracking in urban Shanghai.

[0058] Embodiment 2 of the present invention provides a low-altitude aircraft environmental noise modeling system, comprising: The first quantization module is used to construct a spherical coordinate system of the environment under test, and convert the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system into a sound source directivity tensor. The second quantization module is used to convert the three-dimensional environmental terrain information and three-dimensional building geometric information of the environment under test into a terrain layout tensor. The feature extraction module is used to extract high-level features of the sound source directivity tensor and terrain layout tensor, as well as the frequency and flight altitude features of the low-altitude aircraft at the current moment. The fusion module is used to fuse the high-level features of the sound source directivity tensor and the terrain layout tensor, the frequency features and flight altitude features of the low-altitude aircraft at the current moment, to obtain the multimodal latent features at the current moment. The prediction module is used to obtain the noise intensity map of the low-altitude aircraft at the current moment based on the multimodal potential features at the current moment.

[0059] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0060] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for modeling environmental noise of low-altitude aircraft, characterized in that, include: Construct a spherical coordinate system for the environment under test, and convert the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system into a sound source directivity tensor; The three-dimensional environmental terrain information and three-dimensional building geometry information of the environment under test are converted into a terrain layout tensor. High-level features of the sound source directivity tensor and terrain layout tensor are extracted respectively, as well as the frequency and flight altitude features of the low-altitude aircraft at the current moment. By fusing the high-level features of the sound source directivity tensor and the terrain layout tensor, the frequency features and flight altitude features of the low-altitude aircraft at the current moment, the multimodal latent features at the current moment are obtained. Based on the multimodal potential features at the current moment, obtain the noise intensity map of the low-altitude aircraft at the current moment.

2. The method for modeling environmental noise of low-altitude aircraft according to claim 1, characterized in that, The process of obtaining the multimodal latent features at the current time includes: After concatenating the sound source directivity tensor and the terrain layout tensor, the output features of each encoder layer of the first U-Net model are obtained through the encoder of the first U-Net model. The frequency and altitude of the low-altitude aircraft at the current moment are extracted by passing them through their corresponding learnable embedding layers. The output features of the last encoder layer of the first U-Net model, the frequency features of the low-altitude aircraft at the current moment, and the flight altitude features are fused by element-level addition and then passed through the bottleneck layer of the first U-Net model to obtain the bottleneck layer features of the first U-Net model. The bottleneck layer features of the first U-Net model and the output features of each encoder layer of the first U-Net model are passed through the decoder of the first U-Net model to obtain the multimodal latent features at the current time.

3. The method for modeling environmental noise of low-altitude aircraft according to claim 2, characterized in that, The step of obtaining the noise intensity map of the low-altitude aircraft at the current moment based on the multimodal latent features at the current moment includes: Using the current multimodal latent features as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment; where... This represents the preset total number of time steps.

4. The method for modeling environmental noise of low-altitude aircraft according to claim 3, characterized in that, The current multimodal latent features are used as conditional features of the diffusion model, through... The reverse diffusion process at each time step yields the noise intensity map of the low-altitude aircraft at the current moment, including: A temporal embedding vector for each time step is generated using sinusoidal position encoding; The first Noise intensity map at time step, the first time step The temporal embedding vector at the nth time step and the multimodal latent features at the current time step are obtained through the nth time step. The reverse diffusion process at the nth time step yields the nth time step. The noise intensity map at each time step includes: The first Noise intensity map at time step, the first time step The temporal embedding vector at each time step and the multimodal latent features at the current time step are fused together using element-wise addition to obtain the first... The fused feature maps at each time step; among them... , For time step index, the first The noise intensity map at each time step was obtained by sampling from a standard normal distribution; The first The fused feature map at the nth time step is passed through the encoder of the second U-Net model to obtain the nth time step. The second U-Net model at the first time step The target feature map output by each encoder layer; where... This represents the total number of encoder layers in the second U-Net model. The first The second U-Net model at the first time step The target feature map output by the encoder layer is passed through a multi-head self-attention block to obtain the first... The second U-Net model at the first time step Multi-head self-attention feature maps corresponding to each encoder layer; The first The second U-Net model at the first time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the bottleneck layer of the second U-Net model to obtain the first encoder layer. Bottleneck feature map of the second U-Net model at each time step; The first Bottleneck feature map of the second U-Net model at each time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the decoder of the second U-Net model with residual connections to obtain the first encoder layer. Noise intensity map at each time step; Use the noise intensity map at time step 0 as the noise intensity map of the low-altitude aircraft at the current time.

5. The method for modeling environmental noise of a low-altitude aircraft according to claim 4, characterized in that, The first Bottleneck feature map of the second U-Net model at each time step The multi-head self-attention feature map corresponding to the encoder layer is passed through the decoder of the second U-Net model with residual connections to obtain the first encoder layer. The noise intensity map at each time step includes: The first The bottleneck feature map of the second U-Net model after upsampling at the first time step and the first time step The second U-Net model at the 1st time step The multi-head self-attention feature maps corresponding to the encoder layers are concatenated along the channel dimension and then passed through the first decoder layer of the second U-Net model to obtain the first... Decoding fusion feature map of the first decoder layer of the second U-Net model at each time step; The first The decoding fusion feature map of the first decoder layer of the second U-Net model at the 1st time step, The bottleneck feature map after upsampling the second U-Net model at the nth time step is used to obtain the nth time step through element-wise addition. The target fusion feature map of the first decoder layer of the second U-Net model at each time step; The first The second U-Net model at the 1st time step The target fusion feature map upsampled by the decoder layer and the first decoder layer The second U-Net model at the 1st time step The output of the encoder layer is the first... After the target feature maps at each time step are concatenated along the channel dimension, they are processed by the second U-Net model. The decoder layer is obtained to get the first... The second U-Net model at the 1st time step Decoding fusion feature maps of each decoder layer; where... , For decoder layer index; The first The second U-Net model at the 1st time step Decoding fusion feature map of the decoder layer, the first The second U-Net model at the 1st time step The target fusion feature map upsampled from the decoder layer is used to obtain the th element-wise addition. The second U-Net model at the 1st time step Target fusion feature map of each decoder layer; The first The second U-Net model at the 1st time step The target fusion feature map of the decoder layer is used as the first decoder layer. Noise intensity map at each time step.

6. The method for modeling environmental noise of low-altitude aircraft according to claim 3, characterized in that, An aircraft noise assessment model is constructed based on a multimodal conditional encoder and a diffusion model. The loss function of the aircraft noise assessment model includes: The diffusion denoising loss between the noise intensity map of the low-altitude aircraft at the current moment and the actual noise intensity map; The smoothed L1 loss between the multimodal latent features and the true noise intensity map at the current time.

7. The method for modeling environmental noise of low-altitude aircraft according to claim 1, characterized in that, Sound source directivity data includes: the frequency characteristics and intensity characteristics of the sound.

8. The method for modeling environmental noise of low-altitude aircraft according to claim 1, characterized in that, Construct a spherical coordinate system for the environment under test. Reproject the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system onto a hemisphere, and then convert the hemisphere data into a sound source directivity tensor.

9. The method for modeling environmental noise of low-altitude aircraft according to claim 1, characterized in that, Obtain noise intensity maps of low-altitude aircraft at various times. For each noise intensity map belonging to the same flight altitude, perform noise footprint processing to obtain multiple 2D images belonging to the same flight altitude. Multiple 2D images at the same flight altitude are interpolated to generate a 3D noise distribution visualization.

10. A low-altitude aircraft environmental noise modeling system, characterized in that, include: The first quantization module is used to construct a spherical coordinate system of the environment under test, and convert the sound source directivity data of the low-altitude aircraft at different positions in the spherical coordinate system into a sound source directivity tensor. The second quantization module is used to convert the three-dimensional environmental terrain information and three-dimensional building geometric information of the environment under test into a terrain layout tensor. The feature extraction module is used to extract high-level features of the sound source directivity tensor and terrain layout tensor, as well as the frequency and flight altitude features of the low-altitude aircraft at the current moment. The fusion module is used to fuse the high-level features of the sound source directivity tensor and the terrain layout tensor, the frequency features and flight altitude features of the low-altitude aircraft at the current moment, to obtain the multimodal latent features at the current moment. The prediction module is used to obtain the noise intensity map of the low-altitude aircraft at the current moment based on the multimodal potential features at the current moment.

Images