A method and system for simulating a low-altitude aerial image based on multi-physics field coupling
By employing a multiphysics coupling method, the problem of model simplification in the simulation of low- and medium-altitude aerial images is solved, achieving high-fidelity and efficient image generation, which is applicable to fields such as military reconnaissance and land surveying.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINESE PEOPLES LIBERATION ARMY ARMY SERVICES UNIVERSITY
- Filing Date
- 2026-01-23
- Publication Date
- 2026-06-16
AI Technical Summary
Existing methods for simulating low- and medium-altitude aerial images suffer from oversimplification of models, lack of multi-physics coupling, and failure to accurately simulate complex atmospheric scattering and sensor noise. This results in significant discrepancies between the simulation results and real images in terms of physical consistency and visual realism.
By establishing a two-way coupling relationship between the optical field, the hydrodynamic field, and the sensor physical field, combining Rayleigh scattering and aerosol scattering, introducing a turbulent phase screen model, generating digital simulation images, and constructing a color perspective and contrast attenuation model based on the Beer-Lambert law, multi-physics collaborative calculation is achieved.
It significantly improves the physical realism and visual fidelity of the generated images, adapts to the simulation of high-altitude visual features under various weather conditions, and provides a high-fidelity and efficient data expansion solution, suitable for fields such as military reconnaissance and land surveying.
Smart Images
Figure CN122223583A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer vision and image processing technology, specifically to a method and system for simulating low- and medium-altitude aerial images based on multi-physics coupling. Background Technology
[0002] Unmanned aerial vehicle (UAV) remote sensing and aerial imaging technologies have been widely applied in various fields such as military reconnaissance, geographic surveying and mapping, environmental monitoring, smart agriculture, and emergency rescue. Acquiring high-quality and diverse aerial image data is crucial for driving the development and performance improvement of related algorithms. However, due to limitations in flight costs, airspace control, and complex weather conditions, acquiring a large number of realistic aerial images from low-to-medium altitude fixed-wing UAVs (flight altitude ≥ 1000 meters), especially standard datasets for training and testing intelligent algorithms such as target detection and semantic segmentation, remains a significant challenge. Currently, UAV systems generally lack the ability to perform high-fidelity simulations in complex low-to-medium altitude environments, and current research is largely based on ultra-low altitude rotary-wing UAVs (flight altitude ≥ 1000 meters). The dataset captured at 200 meters (the altitude of the UAV) exhibits significant differences in visual features (such as resolution and atmospheric effects) compared to images obtained from low- and medium-altitude fixed-wing UAV platforms, which limits the generalization performance and robustness of related visual algorithms. Currently, low- and medium-altitude aerial image simulation technology mainly exists in the following development stages and technical routes: (1) Based on simple image processing methods: In the early days, traditional image processing techniques such as color space transformation, contrast reduction, and Gaussian blur were used to simulate the visual effect at high altitudes. However, the images generated by these methods have obvious artificial traces and cannot accurately reflect the changes in real visual characteristics under different altitudes, atmospheric conditions and lighting environments.
[0003] (2) Empirical model-based method: This method uses a preset tone mapping curve to simulate the color characteristics of the upper atmosphere and applies different degrees of blurring based on the estimated depth information. This method relies heavily on manual experience to adjust parameters, has limited generalization ability, and is difficult to adapt to the complex and ever-changing actual atmospheric environment.
[0004] (3) Physical model-based methods: Introducing atmospheric scattering models to consider single scattering events of light in the atmosphere and improving rendering efficiency through pre-computation. However, most methods use oversimplified atmospheric models (such as using only a single scattering coefficient), which cannot accurately simulate the combined effect of Rayleigh scattering and aerosol scattering; at the same time, these methods usually ignore the influence of sensor noise, optical diffraction, signal quantization and other factors in real imaging systems, resulting in significant differences in physical consistency and visual realism between simulated images and real high-altitude images.
[0005] In summary, existing methods for simulating low- and medium-altitude aerial images generally suffer from the following problems: (1) The model is oversimplified, usually only considering a single type of atmospheric scattering and not integrating multiple physics processes; (2) The lack of an effective coupling mechanism between the optical field, the hydrodynamic field and the sensor physical field leads to inconsistencies in the simulation results at the physical level; (3) The simulation of high-altitude image features is relatively simple and lacks the ability to adaptively generate spectral and texture features under diverse weather conditions; (4) Some methods rely on a large amount of real pairing data or complex external inputs, which limits their practicality.
[0006] Therefore, there is an urgent need for a method to simulate low-to-medium altitude aerial images to solve the above problems. Summary of the Invention
[0007] The technical problem to be solved by this invention is how to address the issues of simplified models and lack of multi-physics coupling in existing high-altitude image simulation technology.
[0008] This invention solves the above-mentioned technical problems through the following technical means: a method for simulating low-to-medium altitude aerial images based on multi-physics coupling, comprising: S1. Normalize and convert the data type of the input raw aerial images taken by the ultra-low altitude rotary-wing UAV; S2. Estimate the scene depth information and global atmospheric light value of the original image; S3. Establish a two-way coupling relationship between the optical field, the fluid dynamics field, and the sensor physical field to generate a digital simulation image; S4. Adaptively adjust the spectral characteristics of the digital simulation image, and construct a color perspective and contrast attenuation model based on the Beer-Lambert law to obtain the final simulated visual feature image of the low-to-medium altitude UAV aerial photography.
[0009] Furthermore, in step S2, the estimation strategy for the scene depth information is as follows:
[0010] in, For depth information, For normalized depth values, The normalized depth value located in the central region of the image. The normalized depth value located in the image edge region. For a certain pixel in the image, The angle between the optical axis of the drone camera and the horizontal plane. Center of the image Half of the image diagonal. where is the vertical coordinate of the pixel. This represents the image height.
[0011] Furthermore, in step S2, the estimation of the global atmospheric light value employs an improved dark channel prior algorithm, including: The preprocessed image is converted to a grayscale image using a weighted average, as shown in the expression:
[0012] Among them, R, G, and B are the three primary color channels of the image; For each pixel, the minimum gray value in its local area is taken to generate a dark channel image, and the average gray value of the top 0.1% of the brightest pixels in the dark channel image is selected as the atmospheric light value A.
[0013] Furthermore, S3 includes: The internal coupling of the optical field decomposes atmospheric scattering into two independent physical processes: Rayleigh scattering and aerosol scattering. The corresponding scattering coefficients are calculated by combining the center wavelength of the RGB channel, and the contributions of the two types of scattering are dynamically allocated through a weighting function to generate the original optical field. Optical field-hydrodynamic field coupling is achieved by simulating the modulation effect of atmospheric turbulence on the original optical field using a turbulence phase screen model. The modulated optical field is then mixed with the original optical field based on the turbulence intensity to obtain a hybrid optical field. Multi-physics field-sensor physical field coupling combines the hybrid optical field with the sensor point spread function, photoelectric conversion noise and modulus quantization characteristics to generate a digital simulation image.
[0014] Furthermore, the internal coupling of the optical field specifically includes the following processes: The Rayleigh scattering coefficient and aerosol scattering coefficient of the spectrum are established, with the following expressions:
[0015] in, Rayleigh scattering coefficient, The aerosol scattering coefficient is... The wavelength index for aerosol scattering. Wavelength; Through a weighting function related to aerosol concentration and wavelength By dynamically combining the Rayleigh scattering coefficient and the aerosol scattering coefficient, a composite transmittance is formed. The expression is:
[0016] in, The weight allocation function is calculated using the following formula:
[0017] in, , To adjust the parameters, This refers to the aerosol concentration. The reference wavelength is blue. atmospheric light value and depth information As parameters, establish a spatially and wavelength-dependent atmospheric light intensity function. The expression is:
[0018] in, The angle between the direction of the sun and the direction of observation; Combined with transmittance and atmospheric light intensity function Establish a complete physical equation for atmospheric scattering:
[0019] in, To simulate images in The optical field at that location, The input image pixels.
[0020] Furthermore, the optical field-hydrodynamic field coupling specifically includes the following processes: Constructing a turbulent phase screen, the expression is:
[0021] in, This is random noise after Gaussian filtering. and The corresponding weights; The turbulent-modulated optical field obtained through the turbulent phase screen is expressed as follows:
[0022] The turbulence-modulated optical field is mixed with the original optical field according to the turbulence intensity. The mixing expression is as follows:
[0023] in, To simulate images in Mixed optical field at the location, turbulence intensity Related to atmospheric conditions, the expression is:
[0024] in, This is the normalized proportionality coefficient. The atmospheric refractive index structure constant. This represents the optical transmission path length.
[0025] Furthermore, the calculation formula for the multiphysics-sensor physical field coupling is as follows:
[0026] in, The optical field after multi-physics coupling. For point spread function convolution operation, With a mean of 0 and a variance of Gaussian distributed random noise.
[0027] Furthermore, the specific method for adaptively adjusting the spectral characteristics of the digital simulation image is as follows: To simulate the spectral characteristics under different weather conditions, the RGB channels of the digital simulation image are adjusted in conjunction with physical parameters. The adjustment expression is as follows:
[0028]
[0029] in, , , These are the center wavelengths of the blue, green, and red channels, respectively. It is the adjustment coefficient for the red channel. It is the adjustment coefficient for the blue channel. This is the effect intensity parameter.
[0030] Furthermore, the color perspective and contrast attenuation model constructed based on the Beer-Lambert law is specifically as follows: Based on Beer-Lambert's law, a color perspective model and a contrast attenuation model are constructed, and their calculation formulas are as follows:
[0031] in, and For the adjusted contrast and color, and For the original contrast and color, and This is the attenuation coefficient.
[0032] This invention also provides a mid-to-low altitude aerial image simulation system based on multi-physics coupling, comprising: Image preprocessing module: used to normalize and convert the data types of the input raw aerial images taken by ultra-low altitude rotary-wing UAVs; Scene information estimation module: used to estimate the scene depth information and global atmospheric light value of the original image; Multiphysics coupling modeling module: used to establish the bidirectional coupling relationship between optical field, fluid dynamics field and sensor physical field, and generate digital simulation image; Post-processing module: used to adaptively adjust the spectral characteristics of the digital simulation image, and construct a color perspective and contrast attenuation model based on the Beer-Lambert law to obtain the final simulated visual feature image of low- and medium-altitude UAV aerial photography.
[0033] The advantages of this invention are: (1) This invention proposes a hierarchical, tightly coupled multi-physics collaborative computing framework for simulating aerial images taken by ultra-low-altitude rotary-wing UAVs as images with the visual characteristics of mid-to-low-altitude fixed-wing UAVs. The atmospheric optical field, hydrodynamic field, and sensor physical field are integrated through a mathematical model driven by physical principles. The fields achieve synergistic effects through parameter transfer and mutual modulation (turbulence intensity modulates the optical phase, and sensor parameters determine the final noise level), significantly improving the physical realism and visual fidelity of the generated images.
[0034] (2) This invention improves the traditional atmospheric scattering model by introducing wavelength-adaptive composite scattering and an adjustable aerosol-Rayleigh scattering weighting function. The model can flexibly simulate the high-altitude visual characteristics under various weather conditions, from clear blue skies to hazy gray-yellow, simply by adjusting physical parameters (such as aerosol concentration), and is highly adaptable.
[0035] (3) This invention completes the simulation of physical consistency of the entire link from natural light transmission to digital image generation. It does not rely on a large amount of paired training data, and provides a low-cost, high-efficiency and high-fidelity data expansion scheme for the training and testing of UAV vision algorithms. It is applicable to many fields that require high-altitude images, such as military reconnaissance, land surveying and mapping, and environmental monitoring. Attached Figure Description
[0036] Figure 1 This is a flowchart of a method for simulating low-to-medium altitude aerial images based on multi-physics coupling, according to Embodiment 1 of the present invention. Figure 2 The images show the effects of three depth generation methods (uniform depth, radial depth, and linear depth) according to Embodiment 1 of the present invention; wherein, Figure 2 (a) is a uniform depth map. Figure 2 (b) is the radial depth map. Figure 2 (c) is a linear depth map; Figure 3 The image shown is a simulated aerial photograph taken by a low-to-medium altitude fixed-wing UAV, obtained through simulation according to Embodiment 1 of the present invention; wherein, Figure 3 The first row shows the simulated image under clear weather conditions, and the second row shows the simulated image under hazy weather conditions. Figure 3(a) is an actual aerial image taken by a low-to-medium altitude fixed-wing UAV. Figure 3 (b) is an aerial image taken by a low-altitude rotary-wing UAV. Figure 3 (c) is the image generated after conversion by the present invention; Figure 4 This is a system module flowchart of Embodiment 2 of the present invention. Detailed Implementation
[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0038] Example 1 like Figure 1 As shown, a method for simulating low-to-medium altitude aerial images based on multi-physics coupling includes: S1. Normalize and convert the input raw aerial images taken by the ultra-low altitude rotary-wing UAV.
[0039] Specifically, the input in this embodiment is an RGB image taken by an ultra-low-altitude rotary-wing UAV with a resolution of 1920×1080. The target output is an image simulating a low-to-medium altitude fixed-wing UAV taken at an altitude of over 1000 meters, exhibiting typical characteristics such as a blue tint, reduced contrast, slight blurring, and turbulence effects. The specific steps are as follows: The original image is preprocessed. The input low-altitude rotorcraft drone aerial image is normalized, scaling the pixel values from the original range (e.g., 0-255) to a floating-point range of [0,1]. Furthermore, the image data type is converted to 32-bit floating-point (float32) to improve computational accuracy and stability. This preprocessing stage ensures the image data is within an appropriate numerical range and color space, preparing it for subsequent processing.
[0040] S2. Estimate the scene depth information and global atmospheric light value of the original image.
[0041] Specifically, S2.1 provides three depth information generation strategies: uniform depth, radial depth, and linear depth. For example... Figure 2 As shown, uniform depth is suitable for flat terrain; radial depth is suitable for drone-captured images taken at an angle; and linear depth is suitable for drone-captured images taken at a large angle. Adaptively estimating scene depth information provides key input to the physical model. The depth map estimation strategy formula is as follows:
[0042] in, For normalized depth values, The normalized depth value located in the central region of the image. The normalized depth value located in the image edge region. For a certain pixel in the image, The angle between the drone camera's optical axis and the horizontal plane, and when perpendicularly downwards, is... When horizontal Upward is positive. Center of the image Half of the image diagonal. where is the vertical coordinate of the pixel. This represents the image height. This strategy can adaptively select a uniform, radial, or linear depth mode based on the image content.
[0043] When pitch angle Very small, meaning when the drone camera is vertically downward, using uniform depth (i.e., the first segment function), when the pitch angle... Less than 45 degrees, meaning the drone camera is tilted, using radial depth (i.e., the second segment function), when the pitch angle... For angles greater than 45 degrees, i.e., when the drone camera is at a large tilt angle, linear depth (i.e., the third segment function) is used.
[0044] S2.2, Figure 1 Atmospheric light estimation in the middle atmosphere is based on the brightness characteristics of pixels in an image to estimate the color and intensity of atmospheric light. A weighted average method is used to convert the preprocessed RGB image to a grayscale image, expressed as:
[0045] Calculate the dark channel of a grayscale image: For each pixel, take the minimum grayscale value within its local area (15×15 window) to generate the dark channel image.
[0046] The grayscale values of the top 0.1% of brightest pixels in the dark channel image are selected, and the mean of these pixels is calculated as the atmospheric illumination value A. The atmospheric illumination value A represents the intensity and color of upper-level atmospheric path radiation. The improved dark channel prior uses grayscale images and the mean (instead of the maximum value) to reduce noise interference, improve estimation robustness, and provide global illumination parameters for subsequent atmospheric scattering models.
[0047] S3. Establish a two-way coupling relationship between the optical field, the fluid dynamics field, and the sensor physical field to generate a digital simulation image.
[0048] Specifically, by establishing a bidirectional coupling relationship between the optical field, the hydrodynamic field, and the sensor's physical field, a complete and coherent physical simulation link from scene radiation to digital image is achieved. The physical fields do not act in isolation, but rather collaborate through shared parameters (depth, wavelength) and mutual modulation (such as turbulence-modulated optical transmission, and sensor characteristics affecting the final noise). The specific steps include: S3.1, Coupling within the optical field: A collaborative calculation of an atmospheric physical scattering model using Rayleigh scattering and aerosol scattering.
[0049] S3.1.1 Composite Transmittance Calculation Model. Traditional atmospheric scattering models often use a single scattering coefficient. This invention decomposes it into two independent physical processes: Rayleigh scattering (molecular scattering) and aerosol scattering (particle scattering). The relative intensity of these two scattering processes directly determines the overall color tone of the image: when Rayleigh scattering dominates, the image appears bluish (clear weather); when aerosol scattering dominates, the image is biased towards grayish-white or dark yellow (haze, dust storms). Based on wavelength parameters... The Rayleigh scattering coefficient and aerosol scattering coefficient are calculated. This invention introduces a wavelength parameter. (For RGB channels, use the center wavelength: Blue channel) Green Channel Red Channel ), calculate the Rayleigh scattering coefficient and the aerosol scattering coefficient:
[0050] in, Rayleigh scattering coefficient, The aerosol scattering coefficient is... The wavelength index for aerosol scattering. Different values are adapted to different atmospheric conditions.
[0051] Through a weighting function related to aerosol concentration and wavelength By dynamically combining the Rayleigh scattering coefficient and the aerosol scattering coefficient, a composite transmittance is formed. The expression is:
[0052] in, The weight allocation function is calculated using the following formula:
[0053] in, , To adjust the parameters, This refers to the aerosol concentration. The blue reference wavelength , and (Aerosol concentration) is set according to the simulated atmospheric conditions (sunny weather). Smoggy weather The composite transmittance model decomposes scattering into Rayleigh scattering (dominantly blue tones) and aerosol scattering (dominantly gray-white tones). Through wavelength-dependent weighting, the model can adaptively balance the contributions of the two types of scattering, thereby flexibly generating a variety of high-altitude visual features from clear blue skies to hazy gray-yellow hues. It accurately simulates the composite scattering effects under different atmospheric conditions and avoids the oversimplification of traditional single scattering models.
[0054] S3.1.2, Spatial and wavelength-dependent atmospheric light intensity functions. These are derived from atmospheric light values. and depth information Constructing the atmospheric light intensity function The formula for simulating the spatial and spectral non-uniformity of atmospheric light is as follows:
[0055] in, Atmospheric light value, It is the angle between the direction of the sun and the direction of observation (in this example, the sun's azimuth angle is set to 45 degrees).
[0056] S3.1.3, Complete Atmospheric Scattering Physics Equations. The fundamental physical process of atmospheric scattering is scene radiation. (Input image pixel values) via transmittance Attenuation, plus atmospheric light intensity function Generates a pristine optical field with a fog-like effect. The calculation formula is as follows:
[0057] By applying the above equations to the RGB channels respectively, a preliminary image of atmospheric scattering effect is obtained.
[0058] S3.2 Optical Field-Hydrodynamic Field Coupling: Turbulent Modulated Light Transmission. High-altitude atmospheric turbulence causes random fluctuations in local refractive index, resulting in wavefront distortion. This invention quantifies this effect using a computational fluid dynamics-inspired phase screen model and directly applies it to the optical field generated in step S3.1. This coupling method simulates the dynamic modulation process of turbulence on the imaging optical path in the real world, forming the basis for generating realistic degradation effects such as high-frequency jitter and local blur. The coupling steps are as follows: Figure 1 As shown. First, a turbulent phase screen is generated. Turbulent Phase Screen The expression is:
[0059] in, This is random noise after Gaussian filtering. and The corresponding weights.
[0060] The optical field modulated by turbulence is obtained through a turbulent phase screen, and its expression is:
[0061] The turbulence-modulated optical field is mixed with the original optical field according to the turbulence intensity:
[0062] in, To simulate images in Mixed optical field at the location, turbulence intensity Related to atmospheric conditions, the expression is:
[0063] in, This is the normalized proportionality coefficient. The atmospheric refractive index structure constant. This is the optical transmission path length (approximately the difference between the drone's flight altitude and the ground target's altitude).
[0064] S3.3, Multiphysics-Sensor Physical Field Coupling. This coupling combines the continuous optical radiation field, modulated by preceding atmospheric and turbulence effects, with the physical characteristics of the discretized imaging sensor (including blurring caused by the point spread function of the optical system, noise in the photoelectric conversion process, and quantization of the analog-to-digital conversion). This step ensures that the digital image output from the analog circuit exhibits high consistency with the image acquired by the actual UAV-borne sensor in terms of noise characteristics, sharpness, and dynamic range. The calculation formula is:
[0065] in, The optical field after multi-physics coupling. For point spread function convolution operation, With a mean of 0 and a variance of Gaussian distributed random noise.
[0066] Will Scale to the [0,255] range and convert to an 8-bit unsigned integer (uint8) to simulate the quantization process of a digital camera.
[0067] S4. Adaptively adjust the spectral characteristics of the digital simulation image, and construct a color perspective and contrast attenuation model based on the Beer-Lambert law to obtain the final simulated visual feature image of the low-to-medium altitude UAV aerial photography.
[0068] Specifically, the spectral characteristics of the image are adjusted and atmospheric perspective effects are added to give the image a typical high-altitude visual style. The specific steps are as follows: S4.1 Adaptive Adjustment of Spectral Characteristics. To simulate the spectral characteristics under different weather conditions, the RGB channels of the quantized image generated in step S3 are adjusted in conjunction with the physical parameters: Red channel adjustment:
[0069] Blue channel adjustment:
[0070] in, , , These are the center wavelengths of the blue, green, and red channels, respectively. It is the adjustment coefficient for the red channel. It is the adjustment coefficient for the blue channel. This is a parameter for effect intensity. When simulating clear weather, the blue channel is enhanced (…). ), inhibiting the red channel ( The blue color is typically "blue sky". When simulating smog or sandstorm weather, the parameters can be adjusted accordingly to reduce the blue enhancement effect. The red and green channels can even be specifically adjusted to simulate the dark yellow and grayish-white color characteristics.
[0071] S4.2 Color Perspective and Contrast Attenuation Modeling. As flight altitude increases, due to the increased atmospheric perspective effect, images exhibit color fading and reduced contrast. Based on the Beer-Lambert law, a color perspective model and a contrast attenuation model are constructed as follows:
[0072] in, and For the adjusted contrast and color, and For the original contrast and color, and This is the attenuation coefficient. For example... Figure 3 These are simulated aerial images taken by a low-to-medium altitude fixed-wing UAV after simulation.
[0073] After all the above steps, the input ultra-low-altitude rotary-wing UAV aerial images are converted into simulated images with the visual characteristics of mid-to-low-altitude fixed-wing UAV aerial images. The output images exhibit a blue tint, low contrast, slight blurring, and turbulence effects, and can be used for UAV target detection training data augmentation or virtual reality applications.
[0074] Example 2 Based on implementation 1, embodiment 2 also provides a mid-to-low altitude aerial image simulation system based on multi-physics coupling, such as... Figure 4 As shown, it includes: Image preprocessing module: Used to normalize and convert the data types of the input raw aerial images taken by ultra-low altitude rotary-wing UAVs.
[0075] Scene information estimation module: used to estimate scene depth information and global atmospheric light value of the original image.
[0076] Specifically, the scene depth information estimation unit is used to construct the scene depth information estimation strategy to achieve three depth distributions: uniform, radial, and linear.
[0077] The strategy for estimating scene depth information is as follows:
[0078] in, For depth information, For normalized depth values, The normalized depth value located in the central region of the image. The normalized depth value located in the image edge region. For a certain pixel in the image, The angle between the drone camera's optical axis and the horizontal plane, and when perpendicularly downwards, is... When horizontal Upward is positive. Center of the image Half of the image diagonal. where is the vertical coordinate of the pixel. This represents the image height.
[0079] Global Atmospheric Illumination Estimation Unit: Used to estimate global atmospheric illumination values using an improved dark channel prior algorithm. Specifically, the preprocessed image is converted to a grayscale image via a weighted average, expressed as:
[0080] R, G, and B are the three primary color channels of an image.
[0081] For each pixel, the minimum gray value in its local area is taken to generate a dark channel image. The average gray value of the top 0.1% of the brightest pixels in the dark channel image is selected as the atmospheric light value A.
[0082] Multiphysics coupling modeling module: used to establish the bidirectional coupling relationship between optical field, fluid dynamics field and sensor physical field, and generate digital simulation image.
[0083] Specifically, the internal coupling unit of the optical field decomposes atmospheric scattering into two independent physical processes: Rayleigh scattering and aerosol scattering. It calculates the corresponding scattering coefficients based on the center wavelengths of the RGB channels and dynamically allocates the contributions of the two types of scattering through a weighting function to generate the original optical field. This includes the following processes: Establish the spectrally accurate Rayleigh scattering coefficient and aerosol scattering coefficient, with the following expressions:
[0084] in, Rayleigh scattering coefficient, The aerosol scattering coefficient is... The wavelength index for aerosol scattering. λ is the wavelength.
[0085] Through a weighting function related to aerosol concentration and wavelength By dynamically combining the Rayleigh scattering coefficient and the aerosol scattering coefficient, a composite transmittance is formed. The expression is:
[0086] in, The weight allocation function is calculated using the following formula:
[0087] in, , To adjust the parameters, This refers to the aerosol concentration. The reference wavelength is blue.
[0088] atmospheric light value and depth information As parameters, establish a spatially and wavelength-dependent atmospheric light intensity function. The expression is:
[0089] in, It is the angle between the direction of the sun and the direction of observation.
[0090] Combined with transmittance and atmospheric light intensity function Establish a complete physical equation for atmospheric scattering:
[0091] in, To simulate images in The optical field at that location, The input image pixels.
[0092] Optical field-hydrodynamic field coupling unit: The modulation effect of atmospheric turbulence on the original optical field is simulated using a turbulent phase screen model. The modulated optical field is then mixed with the original optical field based on the turbulence intensity to obtain a hybrid optical field. Specifically, the process includes the following: Constructing a turbulent phase screen, the expression is:
[0093] in, This is random noise after Gaussian filtering. and The corresponding weights.
[0094] The optical field modulated by turbulence is obtained through a turbulent phase screen, and its expression is:
[0095] The turbulence-modulated optical field is mixed with the original optical field according to the turbulence intensity. The mixing expression is as follows:
[0096] in, To simulate images in Mixed optical field at the location, turbulence intensity Related to atmospheric conditions, the expression is:
[0097] in, This is the normalized proportionality coefficient. The atmospheric refractive index structure constant. This is the optical transmission path length (approximately the difference between the drone's flight altitude and the ground target's altitude).
[0098] Multiphysics-sensor physics coupling unit: Combines the hybrid optical field with the sensor point spread function, photoelectric conversion noise and modulus quantization characteristics to generate digital simulation images.
[0099] The formula for calculating the multiphysics-sensor physical field coupling is:
[0100] in, The optical field after multi-physics coupling. For point spread function convolution operation, With a mean of 0 and a variance of Gaussian distributed random noise.
[0101] Post-processing module: Used to adaptively adjust the spectral characteristics of digital simulation images and construct a color perspective and contrast attenuation model based on Beer-Lambert's law to obtain the final simulated visual feature image of low- and medium-altitude UAV aerial photography.
[0102] Specifically, the adaptive spectral characteristic adjustment unit is used to simulate the spectral characteristics under different weather conditions. It adjusts the RGB channels of the digital simulation image in conjunction with physical parameters, and the adjustment expression is as follows:
[0103]
[0104] in, , , These are the center wavelengths of the blue, green, and red channels, respectively. It is the adjustment coefficient for the red channel. It is the adjustment coefficient for the blue channel. This is the effect intensity parameter.
[0105] Color perspective and contrast attenuation modeling unit: Used to construct color perspective and contrast attenuation models based on the Beer-Lambert law. The calculation formula is as follows:
[0106] in, and For the adjusted contrast and color, and For the original contrast and color, and This is the attenuation coefficient.
[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for simulating low-to-medium altitude aerial images based on multiphysics coupling, characterized in that, include: S1. Normalize and convert the data type of the input raw aerial images taken by the ultra-low altitude rotary-wing UAV; S2. Estimate the scene depth information and global atmospheric light value of the original image; S3. Establish a two-way coupling relationship between the optical field, the fluid dynamics field, and the sensor physical field to generate a digital simulation image; S4. Adaptively adjust the spectral characteristics of the digital simulation image, and construct a color perspective and contrast attenuation model based on the Beer-Lambert law to obtain the final simulated visual feature image of the low-to-medium altitude UAV aerial photography.
2. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 1, characterized in that, In step S2, the estimation strategy for the scene depth information is as follows: in, For depth information, For normalized depth values, The normalized depth value located in the central region of the image. The normalized depth value located in the image edge region. For a certain pixel in the image, The angle between the optical axis of the drone camera and the horizontal plane. Center of the image Half of the image diagonal. where is the vertical coordinate of the pixel. This represents the image height.
3. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 1, characterized in that, In step S2, the estimation of the global atmospheric light value employs an improved dark channel prior algorithm, including: The preprocessed image is converted to a grayscale image using a weighted average, as shown in the expression: Among them, R, G, and B are the three primary color channels of the image; For each pixel, the minimum gray value in its local area is taken to generate a dark channel image, and the average gray value of the top 0.1% of the brightest pixels in the dark channel image is selected as the atmospheric light value A.
4. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 1, characterized in that, S3 includes: The internal coupling of the optical field decomposes atmospheric scattering into two independent physical processes: Rayleigh scattering and aerosol scattering. The corresponding scattering coefficients are calculated by combining the center wavelength of the RGB channel, and the contributions of the two types of scattering are dynamically allocated through a weighting function to generate the original optical field. Optical field-hydrodynamic field coupling is achieved by simulating the modulation effect of atmospheric turbulence on the original optical field using a turbulence phase screen model. The modulated optical field is then mixed with the original optical field based on the turbulence intensity to obtain a hybrid optical field. Multi-physics field-sensor physical field coupling combines the hybrid optical field with the sensor point spread function, photoelectric conversion noise and modulus quantization characteristics to generate a digital simulation image.
5. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 4, characterized in that, The internal coupling of the optical field specifically includes the following processes: The Rayleigh scattering coefficient and aerosol scattering coefficient of the spectrum are established, with the following expressions: in, Rayleigh scattering coefficient, The aerosol scattering coefficient is... The wavelength index for aerosol scattering. Wavelength; Through a weighting function related to aerosol concentration and wavelength By dynamically combining the Rayleigh scattering coefficient and the aerosol scattering coefficient, a composite transmittance is formed. The expression is: in, The weight allocation function is calculated using the following formula: in, , To adjust the parameters, This refers to the aerosol concentration. The reference wavelength is blue. atmospheric light value and depth information As parameters, establish a spatially and wavelength-dependent atmospheric light intensity function. The expression is: in, The angle between the direction of the sun and the direction of observation; Combined with transmittance and atmospheric light intensity function Establish a complete physical equation for atmospheric scattering: in, To simulate images in The optical field at that location, The input image pixels.
6. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 4, characterized in that, The optical field-hydrodynamic field coupling specifically includes the following processes: Constructing a turbulent phase screen, the expression is: in, This is random noise after Gaussian filtering. and The corresponding weights; The turbulent-modulated optical field obtained through the turbulent phase screen is expressed as follows: The turbulence-modulated optical field is mixed with the original optical field according to the turbulence intensity. The mixing expression is as follows: in, To simulate images in Mixed optical field at the location, turbulence intensity Related to atmospheric conditions, the expression is: in, This is the normalized proportionality coefficient. The atmospheric refractive index structure constant. This represents the optical transmission path length.
7. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 4, characterized in that, The calculation formula for the multiphysics-sensor physical field coupling is as follows: in, The optical field after multi-physics coupling. For point spread function convolution operation, With a mean of 0 and a variance of Gaussian distributed random noise.
8. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 1, characterized in that, The specific method for adaptively adjusting the spectral characteristics of the digital simulation image is as follows: To simulate the spectral characteristics under different weather conditions, the RGB channels of the digital simulation image are adjusted in conjunction with physical parameters. The adjustment expression is as follows: in, , , These are the center wavelengths of the blue, green, and red channels, respectively. It is the red channel adjustment coefficient. It is the adjustment coefficient for the blue channel. This is the effect intensity parameter.
9. The method for simulating low-to-medium altitude aerial images based on multiphysics coupling according to claim 1, characterized in that, The color perspective and contrast attenuation model constructed based on the Beer-Lambert law is as follows: Based on Beer-Lambert's law, a color perspective model and a contrast attenuation model are constructed, and their calculation formulas are as follows: in, and For the adjusted contrast and color, and For the original contrast and color, and This is the attenuation coefficient.
10. A mid-to-low altitude aerial image simulation system based on multiphysics coupling, characterized in that, include: Image preprocessing module: used to normalize and convert the data types of the input raw aerial images taken by ultra-low altitude rotary-wing UAVs; Scene information estimation module: used to estimate the scene depth information and global atmospheric light value of the original image; Multiphysics coupling modeling module: used to establish the bidirectional coupling relationship between optical field, fluid dynamics field and sensor physical field, and generate digital simulation image; Post-processing module: used to adaptively adjust the spectral characteristics of the digital simulation image, and construct a color perspective and contrast attenuation model based on the Beer-Lambert law to obtain the final simulated visual feature image of low- and medium-altitude UAV aerial photography.