A visibility estimation method based on digital twin technology

Abstract

The application provides a visibility estimation method and system based on a digital twin technology, and relates to the technical field of meteorological detection.The method comprises the following steps: based on the digital twin concept, simulating atmospheric conditions by using a computer, and generating a medium and low visibility image dataset; wherein each data sample in the medium and low visibility image dataset uses visibility as a label; iteratively training a visibility estimation network by using the medium and low visibility image dataset; acquiring a real-time collected image, and inputting the image into the trained visibility estimation network to obtain a visibility estimation result.The application can simulate atmospheric conditions such as smoke by using a computer based on the digital twin concept, can use accurate visibility as a label during the scene building process in the dataset generation stage, and can create a large number of samples with various types and sufficient quantities by using a domain randomization method, so that the accuracy of the trained model is fundamentally ensured.

Description

Technical Field

[0001] This invention belongs to the field of meteorological observation technology, and in particular relates to a visibility estimation method and system based on digital twin technology. Background Technology

[0002] Visibility is generally defined as an indicator describing the farthest distance at which light or objects can be clearly seen by the human eye or sensors. This indicator typically depends on meteorological conditions, air quality, human visual acuity, or sensor performance. In meteorology, it is generally used to reflect the degree of atmospheric turbidity, and is measured in meters or kilometers. Visibility is widely used in transportation, outdoor operations, atmospheric science, and environmental monitoring, and its timeliness, accuracy, and versatility have a significant impact on disaster early warning, safety assurance, and decision-making. Existing measurement methods generally rely on specialized hardware equipment, which is costly to produce and maintain, requires frequent calibration, and its lifespan and accuracy are affected by harsh operating conditions such as temperature and humidity.

[0003] Atmospheric visibility can generally be measured visually, or using instruments such as atmospheric transilluminators and automatic laser visibility meters. Currently, visibility observations are mostly based on manual visual estimation, which is relatively less standardized and objective. Atmospheric transilluminators measure the transmittance of an atmospheric column by passing a beam of light through two fixed points, thus estimating visibility. This method requires the beam to pass through a sufficiently long atmospheric column, and the reliability of the measurement is affected by the stability of the light source and other hardware systems. It is generally only suitable for observations of moderate to low visibility. In low visibility weather such as rain and fog, complex conditions such as water vapor absorption can cause significant errors. Automatic laser visibility meters estimate visibility by measuring the atmospheric extinction coefficient with a laser. Relatively speaking, this method is more objective and accurate. However, these instruments are expensive, have high maintenance costs, and are complex to operate. Moreover, they are difficult to use in rainy or foggy weather, thus hindering their widespread adoption.

[0004] Machine learning is an effective method to circumvent hardware dependencies. With the continuous development of machine learning, neural networks have made significant improvements in reasoning capabilities, especially in the field of computer vision. Image classification is one of the most fundamental and mature tasks in this area, and with adaptation, it can serve as a basic model for visibility estimation. By fitting images to visibility data, neural networks can infer the average visibility of the surrounding environment in that direction in real time based on the acquired images (hereinafter referred to as "visibility estimation"). The nature of this task dictates that using traditional methods to construct datasets may face problems such as stringent technical requirements, high economic costs, and long production cycles. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a visibility estimation method and system based on digital twin technology.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The first aspect of this invention discloses a visibility estimation method based on digital twin technology, the method comprising:

[0008] Step S1: Based on the concept of digital twin, atmospheric conditions are simulated by computer, and a dataset of medium and low visibility images is generated; wherein, each data sample in the medium and low visibility image dataset uses visibility as the label.

[0009] Step S2: Iteratively train the visibility estimation network using a dataset of medium-low visibility images;

[0010] Step S3: Acquire real-time images and input them into the trained visibility estimation network to obtain visibility estimation results.

[0011] In step S1, based on the concept of digital twins, atmospheric conditions are simulated by computer, and a dataset of low-to-medium visibility images is generated, specifically including:

[0012] S11, initialize the scene and import the basic components that constitute the scene, and then randomly set the parameters of the basic components that constitute the scene based on the principle of domain randomization to obtain depth maps under different scenes; for each depth map under a scene, execute step S12.

[0013] S12, set volumetric fog parameters to generate volumetric fog, and then render the depth map under the volumetric fog according to a certain step size to obtain medium and low visibility images of the current scene under different visibility conditions, and obtain the visibility corresponding to the medium and low visibility image based on the visibility fitting function to generate visibility labels.

[0014] S13. Repeat steps S11-S12 until a dataset of medium-low visibility images that meets the quantity requirements is obtained.

[0015] In step S11, the basic components constituting the scene include a digital twin simulation model, textures, light sources, particle emitters, physical fields, and a virtual camera.

[0016] Step S12, the specific steps for obtaining the visibility fitting function include:

[0017] S121, Initialize the scene and import the reference object and volumetric fog;

[0018] S122, Adjust the volumetric fog parameters to obtain the visible boundary, and calculate the depth value of the visible boundary;

[0019] S123, repeat steps S121-S122 until the parameters of volumetric fog in S122 are completely traversed within the specified range to obtain a parameter-visibility table under arbitrary volumetric fog parameters, and then obtain the visibility fitting function based on the parameter-visibility table.

[0020] In step S123, based on the parameter-visibility table, a hash query or multivariate function fitting is performed to obtain the visibility fitting function, i.e.:

[0021] D vis =f(ρ,i e ,r,g,b)

[0022] Among them, D vis For visibility, ρ, i e r, g, and b are volumetric fog parameters, corresponding to concentration, luminescence intensity, and RGB values, respectively.

[0023] A second aspect of this invention discloses a visibility estimation system based on digital twin technology, comprising:

[0024] The first processing module is configured to simulate atmospheric conditions using a computer based on the concept of digital twins and generate a dataset of medium- and low-visibility images; wherein each data sample in the medium- and low-visibility image dataset uses visibility as a label.

[0025] The second processing module is configured to iteratively train the visibility estimation network using a low-to-medium visibility image dataset.

[0026] The third processing module is configured to acquire real-time images and input them into the trained visibility estimation network to obtain visibility estimation results.

[0027] Based on the concept of digital twins, atmospheric conditions are simulated by computer to generate a dataset of medium- and low-visibility images, specifically including:

[0028] The scene is initialized and its basic components are imported. Then, based on the principle of domain randomization, the parameters of the basic components are randomly set to obtain depth maps for different scenes. For each scene's depth map, the next process is executed:

[0029] Set volumetric fog parameters to generate volumetric fog, and then render the depth map under the volumetric fog according to a certain step size to obtain medium and low visibility images of the current scene under different visibility conditions. Based on the visibility fitting function, obtain the visibility corresponding to the medium and low visibility image to generate visibility labels.

[0030] Repeat the above process of obtaining depth maps and generating visibility labels until a dataset of medium-low visibility images that meets the quantity requirements is obtained.

[0031] The basic components that make up the scene include digital twin simulation models, textures, light sources, particle emitters, physical fields, and virtual cameras.

[0032] A third aspect of this invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of a visibility estimation method based on digital twin technology according to any one of the first aspects of this disclosure.

[0033] A fourth aspect of this invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a visibility estimation method based on digital twin technology according to any one of the first aspects of this disclosure.

[0034] In summary, the solution proposed in this invention has the following technical effects:

[0035] Improved robustness under complex conditions: This invention uses visible light images as input, and the dataset simulates scenes under various complex lighting conditions. Its accuracy is almost unaffected by complex factors such as atmospheric turbidity, atmospheric composition, and color difference.

[0036] Improved visibility estimation accuracy: Based on the concept of digital twins, this invention uses computer simulation of atmospheric conditions such as smoke to accurately label visibility during the scene construction process in the dataset generation stage. Furthermore, it employs domain randomization to create diverse and abundant samples, thereby fundamentally ensuring the accuracy of the trained model.

[0037] Reduced economic costs and operational difficulty: This invention eliminates the need for expensive sensors and hardware devices, using common image acquisition devices as input sources (such as digital cameras). The core program can run on most general-purpose computers and is user-friendly, requiring no professional training in the operation and maintenance of the instrument. Attached Figure Description

[0038] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0039] Figure 1This is a flowchart of a visibility estimation method based on digital twin technology according to an embodiment of the present invention;

[0040] Figure 2 A flowchart illustrating the batch generation of datasets according to an embodiment of the present invention;

[0041] Figure 3 This is a flowchart illustrating the generation of a fitted visibility function according to an embodiment of the present invention;

[0042] Figure 4 This is a structural diagram of a visibility estimation system based on digital twin technology according to an embodiment of the present invention;

[0043] Figure 5 This is a structural diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0044] 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 with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and 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.

[0045] It is understood that the terms "first," "second," etc., used herein may be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of this application, a first image may be referred to as a second image, and similarly, a second image may be referred to as a first image. Both the first image and the second image are images, but they are not the same image.

[0046] Digital twins are a simulation concept that has matured in recent years, representing objects, systems, or processes in the physical world through computer simulation. This concept aims to accurately reproduce corresponding entities in the real world virtually and update them in real time. In the field of computer vision, one of the basic methods of digital twins is to build "photorealistic" simulation datasets by using 3D modeling based on computer graphics principles, rendering images, and generating labels. This technology effectively reduces data acquisition costs, improves acquisition efficiency, and solves the problem of some scenes being difficult to create manually in reality. More importantly, the scene information in the computer simulation can be used to directly label the dataset samples, effectively reducing labor costs and improving label accuracy.

[0047] According to an embodiment of the present invention, in a first aspect, a visibility estimation method based on digital twin technology is provided, see [link to relevant documentation]. Figure 1 The method includes:

[0048] Step S1: Based on the concept of digital twin, atmospheric conditions are simulated by computer, and a dataset of medium and low visibility images is generated; wherein, each data sample in the medium and low visibility image dataset uses visibility as the label.

[0049] Please see Figure 2 In step S1, based on the concept of digital twins, atmospheric conditions are simulated by computer, and a dataset of low-to-medium visibility images is generated, specifically including:

[0050] S11, initialize the scene and import the basic components that constitute the scene, and then randomly set the parameters of the basic components that constitute the scene based on the principle of domain randomization to obtain depth maps under different scenes; for each depth map under a scene, execute step S12.

[0051] In step S11, the basic components constituting the scene include a digital twin simulation model, textures, light sources, particle emitters, physical fields, and a virtual camera.

[0052] A digital twin simulation model, also known as a "3D model," is a data structure used to represent three-dimensional objects. It provides 3D modeling software with shape, texture, normals, and other information. Textures further enhance the model's realism, defining information beyond the vertices on each triangle face, including base color, roughness, metallicity, and normal direction. Light sources simulate different lighting intensities. Particle emitters generate particles in the 3D modeling software; these particles are typically used to simulate various special effects, including flames, smoke, and fluids. Physics fields are tools in 3D modeling software used to control the physical simulation of specified objects and particles within a scene. They can interact with objects in forms such as neutrality, magnetism, wind, eddies, and orbital constraints.

[0053] S12, set volumetric fog parameters to generate volumetric fog, and then render the depth map under the volumetric fog according to a certain step size to obtain medium and low visibility images of the current scene under different visibility conditions. Based on the visibility fitting function, obtain the visibility corresponding to the medium and low visibility image to generate a visibility label. The visibility label serves as the image representation of a certain basic scene under different visibility conditions.

[0054] By calculating the deviation scores for the volumetric fog parameters in formula (1) below, we can obtain the influence law of visibility on these variables, and use this as a basis to set the volumetric fog parameters to achieve linear changes in visibility. Since the correlation between the RGB gradient of the pixels near the "visible boundary" and the changes in visibility, as well as the correlation between visibility and the changes in volumetric fog parameters, are differentiable, we can choose a relatively large step size and then use the difference method to set the parameters.

[0055] Please see Figure 3 In step S12, the specific steps for obtaining the visibility fitting function include:

[0056] S121, Initialize the scene and import the reference object and volumetric fog;

[0057] S122, Adjust the volumetric fog parameters to obtain the visible boundary, and calculate the depth value of the visible boundary;

[0058] S123, repeat steps S121-S122 until the parameters of volumetric fog in S122 are completely traversed within the specified range to obtain a parameter-visibility table under arbitrary volumetric fog parameters, and then obtain the visibility fitting function based on the parameter-visibility table.

[0059] In an example scenario, the example object is placed perpendicular to the virtual camera lens, and the side of the example object is exposed to the camera lens to generate a depth map. Volumetric fog parameters are set according to a certain step size to generate volumetric fog, and then the depth map under the volumetric fog is rendered. In the rendering result, the RGB difference of adjacent pixels belonging to the example object is calculated along the direction of the example object, and pixels with adjacent pixel RGB difference below a certain threshold are taken as visible boundaries. The values ​​of nearby pixels less than a distance threshold on the depth map are calculated to infer the depth of the visible boundary.

[0060] In step S124, based on the corresponding table of visibility values ​​under arbitrary volumetric fog parameters, a hash query or multivariate function fitting is performed to obtain the visibility fitting function, i.e.:

[0061] D vis =f(ρ,i e ,r,g,b) (1)

[0062] Among them, D vis For visibility, ρ, i e r, g, and b are volumetric fog parameters, corresponding to concentration, luminescence intensity, and RGB values, respectively.

[0063] It should be noted that the step of obtaining the visibility fitting function is performed only once in the initial process.

[0064] It's important to note that for depth values ​​that are non-linearly compressed along the pixel direction due to depth of field, a method can be used to change the lens's field of view and focal length. Non-linear compression here refers to the fact that, due to the principle of near objects appearing larger and farther objects smaller, in regions with significant depth, a single pixel on the depth map contains a series of points with extremely large depth variations. Such points are difficult to effectively quantize using the default 8-bit integer of a grayscale image. However, while maintaining the same image resolution, changing the field of view and focal length can effectively alter the dynamic range of these points, thereby enabling quantization to be performed to the greatest extent possible in a localized area, reducing rounding errors.

[0065] S13. Repeat steps S11-S12 until a dataset of medium-low visibility images that meets the quantity requirements is obtained.

[0066] In each iteration, the previous scene is cleared, and all models, textures, lights, particle emitters, physics fields, virtual cameras, etc., used for building the current scene are imported as scene assets. Based on the principle of domain randomization, the position, pose, size, and other parameters of these assets are randomly set to ultimately form a fairly general scene. This step is a fundamental prerequisite for ensuring that the feature distribution of the rendering result is consistent with reality. In addition, heterogeneous and randomized scene design is also key to a rich and diverse dataset.

[0067] Considering the impact of image acquisition equipment on visibility, it is necessary to manually adapt the virtual camera's lens parameters and rendering settings based on the fundamental principles of different rendering engines. Lens parameters include lens size, resolution, depth of field, field of view, focal length, aperture, and lens distortion. Rendering settings include the image's color space, color jitter, motion blur, and exposure.

[0068] This step repeats steps S11-S12 to perform a series of visibility image renderings under different scene designs. The purpose of this step is to increase the variety and number of scenes, allowing the neural network to obtain more comprehensive and richer information, thereby improving the performance of its feature extraction module. In randomization methods, it is generally considered that the number of scene types should be above 8,000; provided the training method and hardware support it, the more the better.

[0069] Step S2: Iteratively train the visibility estimation network using a dataset of medium-low visibility images;

[0070] In this step, the dataset constructed in step four is used for training and testing the basic image classification network for visibility estimation, with visibility as the classification label, and the process is iterated until the neural network test results reach the expected accuracy.

[0071] Since the relationship between visibility and pixel changes in the same scene is a continuous sequence, it is differentiable compared to text labels, making it more favorable for gradient descent. Theoretically, it can achieve better training results than general image classification networks.

[0072] Step S3: Acquire real-time images and input them into the trained visibility estimation network to obtain visibility estimation results.

[0073] This invention provides a visibility estimation method based on digital twins. Based on the concept of digital twins, this method uses 3D modeling and simulation to construct scenes and generate a dataset of low-to-medium visibility images. The method creates different weather conditions by constructing virtual scenes and parameterizing volumetric fog generation, then performs visibility calibration and image rendering. The visibility calibration is based on the quantitative relationship between volumetric fog parameters and visibility, including parameters such as volumetric fog concentration, luminous intensity, and RGB values. Under the same volumetric fog parameters, the obtained visibility is also the same.

[0074] Because the parameters are manually adjustable, this method can generate images with arbitrary parameters and visibility levels. Furthermore, the scene can be customized by changing the positions of entities such as models, light sources, and virtual cameras, ensuring photorealistic images while effectively increasing the variety and quantity of data. This dataset, used for training a visibility estimation network, effectively improves its accuracy and generalization ability, enabling the neural network to accurately estimate visibility in all weather conditions and environments using only general-purpose image acquisition equipment.

[0075] The second aspect of this invention discloses a visibility estimation system based on digital twin technology. Please refer to [link to relevant documentation]. Figure 4 The system includes:

[0076] The first processing module 100 is configured to simulate atmospheric conditions using a computer based on the concept of digital twins and generate a dataset of medium and low visibility images; wherein each data sample in the medium and low visibility image dataset uses visibility as a label.

[0077] The second processing module 200 is configured to iteratively train the visibility estimation network using a low-to-medium visibility image dataset.

[0078] The third processing module 300 is configured to acquire real-time images and input them into the trained visibility estimation network to obtain visibility estimation results.

[0079] Based on the concept of digital twins, atmospheric conditions are simulated by computer to generate a dataset of medium- and low-visibility images, specifically including:

[0080] The scene is initialized and its basic components are imported. Then, based on the principle of domain randomization, the parameters of the basic components are randomly set to obtain depth maps for different scenes. For each scene's depth map, the next process is executed:

[0081] Set volumetric fog parameters to generate volumetric fog, and then render the depth map under the volumetric fog according to a certain step size to obtain medium and low visibility images of the current scene under different visibility conditions. Based on the visibility fitting function, obtain the visibility corresponding to the medium and low visibility image to generate visibility labels.

[0082] Repeat the above process of obtaining depth maps and generating visibility labels until a dataset of medium-low visibility images that meets the quantity requirements is obtained.

[0083] The basic components that make up the scene include digital twin simulation models, textures, light sources, particle emitters, physical fields, and virtual cameras.

[0084] A third aspect of this invention discloses an electronic device. The electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of a visibility estimation method based on digital twin technology according to any one of the first aspects of this disclosure.

[0085] Figure 5 This is a structural diagram of an electronic device according to an embodiment of the present invention, such as... Figure 5 As shown, the electronic device includes a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, Near Field Communication (NFC), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the device's casing, or an external keyboard, touchpad, or mouse.

[0086] Those skilled in the art will understand that Figure 5 The structure shown is merely a structural diagram of the part related to the technical solution of this disclosure and does not constitute a limitation on the electronic device to which the solution of this application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0087] A fourth aspect of this invention discloses a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a visibility estimation method based on digital twin technology according to any one of the first aspects of this disclosure.

[0088] 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 or all of the technical features therein, and such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A visibility estimation method based on digital twin technology, characterized in that, The method comprises: Step S1, based on the digital twin concept, simulating atmospheric conditions by computer, and generating a low-visibility image dataset; wherein each data sample in the low-visibility image dataset uses visibility as a label; In the step S1, based on the digital twin concept, simulating atmospheric conditions by computer, and generating a low-visibility image dataset, specifically comprising: S11, initialize the scene and import the basic components constituting the scene, and then randomly set the parameters of the basic components constituting the scene based on the domain randomization principle to obtain depth maps under different scenes; for the depth map under each scene, execute step S12; S12, set the volume fog parameters to generate volume fog, and then perform image rendering on the depth map under the volume fog according to a certain step size to obtain low-visibility images of the current scene under different visibilities, and obtain the visibility corresponding to the low-visibility image based on the visibility fitting function to generate the visibility label; In step S12, the step of obtaining the visibility fitting function specifically comprises: S121, initialize the scene and import the reference object and volume fog; S122, adjust the volume fog parameters to obtain the visible boundary and calculate the depth value of the visible boundary; S123, repeat steps S121-S122 until the parameters of the volume fog in S122 are completely traversed within a specified range to obtain a parameter-visibility table under any volume fog parameter, and then obtain the visibility fitting function based on the parameter-visibility table; In step S123, based on the parameter-visibility table, a hash query or a multi-variate function fitting is performed to obtain a visibility fitting function, i.e., , wherein, is the visibility, , , , , is the volume fog parameter, respectively corresponding to the concentration, the luminous intensity and the RGB three-element value; S13, repeating steps S11-S12 until the low-visibility image dataset meeting the quantity requirement is obtained; Step S2, iteratively train the visibility estimation network using the low-visibility image dataset; Step S3, acquire a real-time visible light image and input it into the trained visibility estimation network to obtain a visibility estimation result.

2. The method of claim 1, wherein, In the step S11, the basic components constituting the scene include a digital twin simulation model, a map, a light source, a particle emitter, a physical field, and a virtual camera.

3. A visibility estimation system based on digital twin technology, characterized in that, Comprise: A first processing module is configured to simulate atmospheric conditions by computer based on the digital twin concept, and generate a low-visibility image dataset; wherein each data sample in the low-visibility image dataset uses visibility as a label; Based on the digital twin concept, simulating atmospheric conditions by computer, and generating a low-visibility image dataset, specifically comprising: Initialize the scene and import the basic components constituting the scene, and then randomly set the parameters of the basic components constituting the scene based on the domain randomization principle to obtain depth maps under different scenes; for the depth map under each scene, execute the next process, that is: Set the volume fog parameters to generate volume fog, and then perform image rendering on the depth map under the volume fog according to a certain step size to obtain low-visibility images of the current scene under different visibilities, and obtain the visibility corresponding to the low-visibility image based on the visibility fitting function to generate the visibility label; The above process of obtaining the depth map and generating the visibility label is performed multiple times until a low-visibility image dataset meeting the quantity requirement is obtained; The step of obtaining the visibility fitting function specifically comprises: Initialize the scene and import the reference object and volume fog; Adjusting volume fog parameters, obtaining a visible boundary, and calculating a depth value of the visible boundary; Repeating the scene initialization and volume fog parameter adjustment process until the parameters of the volume fog are completely traversed in the specified range to obtain a parameter-visibility table under any volume fog parameter, and then obtaining a visibility fitting function based on the parameter-visibility table; Based on the parameter-visibility table, a hash query or a multivariate function fitting is performed to obtain a visibility fitting function, i.e.: , wherein, is the visibility, , , , , is the volume fog parameter, respectively corresponding to the concentration, the luminous intensity and the RGB three-element value; The second processing module is configured to iteratively train the visibility estimation network using the low-mid visibility image dataset; The third processing module is configured to obtain a real-time visible light image and input the real-time visible light image into the trained visibility estimation network to obtain a visibility estimation result.

4. The system of claim 3, wherein, The basic components of the scene include a digital twin simulation model, a map, a light source, a particle emitter, a physical field, and a virtual camera.

5. An electronic device, comprising: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of the visibility estimation method based on the digital twin technology in any one of claims 1-2.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the visibility estimation method based on the digital twin technology in any one of claims 1-2.

Images