Rendering engine-based method for enhancing sense of reality of digital twin model

Abstract

The invention relates to the technical field of image enhancement rendering, in particular to a digital twin model reality sense enhancement method based on a rendering engine, and the method comprises the steps: obtaining a rendering image for virtual scene rendering, obtaining a rendering gray image, constructing a texture trend pair of each pixel point, and selecting a trend distribution pair of each pixel point; analyzing the difference between the trend distribution pair of each pixel point and the texture trend pair of each pixel point in a local neighborhood to obtain a structure distribution characteristic degree, and based on clustering division of all pixel points, analyzing the difference condition of the texture trend pairs of all pixel points in a clustering cluster where each pixel point is located to obtain the adjacent structure difference degree of each pixel point; and obtaining the rendering ambiguity of each pixel point by combining the dispersion degree of the gray value of the pixel point in the local neighborhood of each pixel point, obtaining a filtered image by adopting a guided graph filtering algorithm, and rendering the filtered image. The virtual scene rendering effect can be improved, and the rendering authenticity is enhanced.

Description

Technical Field

[0001] This application relates to the technical field of image enhancement rendering, and particularly to a method for enhancing the realism of a digital twin model based on a rendering engine. Background Art

[0002] Digital twin technology is a technology that maps the state of a physical entity through a virtual model. In recent years, with the development of technologies such as big data, cloud computing, the Internet of Things, and artificial intelligence, it has gradually been widely applied in fields such as manufacturing, smart cities, healthcare, and agricultural technology. The core of digital twin technology lies in reflecting the state and operation of a physical entity into a virtual model through real-time data collection and analysis, so as to achieve simulation, prediction, and optimization.

[0003] However, the realization of a digital twin model is inseparable from high-quality visual presentation, and the rendering engine is the key technology to achieve this goal. With the expansion of digital twin application scenarios, higher requirements are put forward for the rendering engine. How to enhance the realism of a digital twin model through the rendering engine has become an important direction for the development of current digital twin technology.

[0004] During the rendering process of a virtual model, traditional technologies generally preprocess images through Guided Image Filtering. However, since traditional Guided Image Filtering generally processes images using a fixed truncation value ε, a too large value will cause image blurring and poor rendering effects, while a smaller truncation value will result in a slower rendering speed. Since virtual technology is real-time, there are differences in images at each moment. The fixed truncation value will cause a conflict between the real-time performance and image quality requirements of virtual model rendering, resulting in problems such as slow rendering speed and weak rendering authenticity. Summary of the Invention

[0005] In order to solve the above technical problems, the purpose of this application is to provide a method for enhancing the realism of a digital twin model based on a rendering engine, and the specific technical solution adopted is as follows: An embodiment of this application provides a method for enhancing the realism of a digital twin model based on a rendering engine, including the following steps: Obtain a rendering image for virtual scene rendering, and obtain a rendered grayscale image through grayscale processing; Construct a texture trend pair for each pixel point based on the distribution characteristics of the grayscale values of the pixel points in the local neighborhood of each pixel point in the rendered grayscale image, and select a trend distribution pair for each pixel point according to the grayscale difference between each pixel point and the local neighborhood pixel points; Analyze the difference between the trend distribution pairs of each pixel and the texture trend pairs of each pixel in the local neighborhood to obtain the structural distribution feature degree between each pixel and each pixel in the local neighborhood, and perform clustering division on all pixels based on the structural distribution feature degree. Obtain the adjacent structure difference degree of each pixel through the difference situation of the texture trend pairs of all pixels in the cluster where each pixel is located after clustering division; Analyze the dispersion degree of the gray values of the pixels in the local neighborhood of each pixel, and combine the adjacent structure difference degree of each pixel to obtain the rendering blur degree of each pixel, and perform linear mapping on the rendering blur degrees of all pixels to obtain the corresponding mapping values of each pixel. Use the guided image filtering algorithm to filter the rendered image and perform rendering processing on the filtered image.

[0006] Preferably, the construction of the texture trend pairs of each pixel includes: for each pixel, respectively form the texture trend pairs of each pixel by combining the gray values of each pixel and its eight-neighborhood pixels.

[0007] Preferably, the determination of the trend distribution pairs of each pixel includes: calculate the absolute value of the gray value difference between each pixel and its eight-neighborhood pixels, and select the texture trend pair corresponding to the neighborhood pixel with the largest absolute value as the trend distribution pair of each pixel.

[0008] Preferably, the corresponding calculation method of the structural distribution feature degree between each pixel and each pixel in the local neighborhood is: Calculate the difference between the trend distribution pair of each pixel and each texture trend pair of the pixels in the eight-neighborhood to obtain the structural direction distribution degree between the trend distribution pair of each pixel and each texture trend pair of each pixel in the eight-neighborhood; Then the expression of the structural distribution feature degree between each pixel and each pixel in the local neighborhood is: ; represents the structural distribution feature degree between pixel u and the i-th pixel in its eight-neighborhood; represents the set composed of the structural direction distribution degrees between the trend distribution pair of pixel u and all texture trend pairs of the i-th pixel in its eight-neighborhood; min() represents the minimum value function.

[0009] Preferably, the corresponding calculation formula of the structural direction distribution degree between the trend distribution pair of each pixel and each texture trend pair of each pixel in the eight-neighborhood is: ; In the formula, represents the structural direction distribution degree between the trend distribution pair of pixel u and the j-th texture trend pair of the i-th pixel in its eight-neighborhood, 、 respectively represent the first and second elements in the trend distribution pair of pixel u; , represent the first and second elements in the j-th texture trend pair of the i-th pixel within the eight-neighborhood of pixel u; || represents the absolute value.

[0010] Preferably, during the process of clustering and partitioning all pixel points, it further includes: forming the structural distribution feature sequence of each pixel point from the structural distribution feature degrees between each pixel point and all pixel points within its eight-neighborhood, using the structural distribution feature sequences of all pixel points as the input of the clustering algorithm, and using the difference in the structural distribution feature degrees between any two pixel points as the metric distance in the clustering process to perform clustering and partitioning on the pixel points.

[0011] Preferably, the calculation method for the proximity structure difference degree of each pixel point is: ; where represents the proximity structure difference degree of pixel u; N represents the number of texture trend pairs in the texture trend set of pixel u; , respectively represent the first and second elements of the k-th texture trend pair in the texture trend set of pixel u; wherein, the texture trend set is the set composed of the texture trend pairs of all pixel points within the cluster where the pixel is located.

[0012] Preferably, the corresponding calculation method for the rendering blur degree of each pixel point is: ; where represents the rendering blur degree of pixel u; represents the standard deviation of the gray values of all pixel points within the eight-neighborhood of pixel u; represents the proximity structure difference degree of pixel u; Norm() is the normalization function; represents a constant to avoid a zero denominator.

[0013] Preferably, further including in the process of filtering the rendered image using the guided image filtering algorithm: using the mapping value corresponding to each pixel point as the truncation value during the process of filtering the rendered image by the guided filtering algorithm to perform filtering processing on each pixel point in the rendered image.

[0014] Preferably, the gray scale processing uses the weighted average method.

[0015] As can be seen from the above, a method for enhancing the realism of a digital twin model based on a rendering engine provided by this application has at least the following beneficial effects: In view of the problems in the prior art that due to the requirements of real-time rendering and image quality of virtual models, the rendering speed is slow and the image quality is not high, the present application analyzes the local features and structural distribution features of pixel points in the virtual scene image in real time to characterize the structural features of different objects in the virtual scene; by calculating the structural distribution feature degree and adjacent structure difference degree of pixel points, it characterizes the boundaries and regions of objects in the virtual scene, and is used to identify the junction situation between different texture structures in the image; Furthermore, the rendering blur degree of pixel points is obtained to determine the blur degree of different positions in the rendered image, so as to maintain the rendering image rate while ensuring the details of rendering, making the rendering effect better. Therefore, the present application avoids the problems of low rendering efficiency and weak rendering authenticity caused by the requirements of real-time performance and image quality in the virtual model rendering method in the prior art, and can effectively improve the rendering quality and efficiency of the virtual model, making the rendering speed of the virtual model faster, the rendering effect better, and the authenticity stronger. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the steps of a method for enhancing the realism of a digital twin model based on a rendering engine provided by the present application; Figure 2 It is a schematic diagram of constructing a texture trend pair for the eight-neighborhood of pixel points provided by the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to further elaborate on the technical means and effects adopted by the present application to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features and effects of a method for enhancing the realism of a digital twin model based on a rendering engine proposed by the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise specified or limited, terms such as "including", "comprising" or any other variant thereof are intended to cover non-exclusive inclusion, such that a circuit structure, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such article or device. Without further limitation, an element defined by the statement "including one..." does not exclude the presence of additional identical elements in the article or device including said element. Additionally, the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. All technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which this application pertains.

[0020] The following specifically describes the specific solution of a method for enhancing the realism of a digital twin model based on a rendering engine provided by this application in conjunction with the accompanying drawings.

[0021] Please refer to Figure 1 , which shows a flowchart of the steps of a method for enhancing the realism of a digital twin model based on a rendering engine provided by an embodiment of this application, including the following steps: Step 1: Obtain a rendering image for virtual scene rendering, and obtain a rendered grayscale image through grayscale processing.

[0022] First, the real-world scene of the digital twin model is tracked in real time through infrared or ultrasonic technology, and the video of the digital twin model is collected by a camera to obtain the video frame image of the digital twin model, denoted as the rendering image. To improve the efficiency of image data processing, the rendering image will be preprocessed. There are many existing methods for image preprocessing, and in the actual application scenario, the implementer can set and select by himself. In this embodiment, the rendering image is converted into a grayscale image using the weighted average method, denoted as the rendered grayscale image. Among them, the calculation of the weighted average method is a well-known technology, and the specific calculation process will not be elaborated here.

[0023] Step 2: Construct the texture trend pairs of each pixel point based on the distribution characteristics of the pixel point gray values in the local neighborhood of each pixel point in the rendered grayscale image, and select the trend distribution pairs of each pixel point according to the gray difference between each pixel point and the local neighborhood pixel points.

[0024] Since in the rendering of virtual scenes, in order to maintain a high rendering effect, it is necessary to maintain high-detail attributes of the scene rendering image. Since various substances in reality do not exist in the rendering image as single pixel points, the structure of the same substance usually has a certain area ratio. The distribution of pixel points within the same substance area has the characteristic of similarity, and the intersections of different structures also have similar characteristics.

[0025] At the boundary junction of different structures, the similarity of the pixels at the junction does not change with the position of the junction. Similar features must have connected features. Thus, in this embodiment, with each pixel of the rendered grayscale image as the center, the eight-neighborhood pixels of the central pixel are obtained, and the grayscale values of the central pixel and its respective eight-neighborhood pixels are respectively formed into texture trend pairs of the central pixel, and the eight texture trend pairs are formed into a texture trend pair set.

[0026] Specifically, in this embodiment, the schematic diagram of the eight-neighborhood distribution of pixels for constructing texture trend pairs is as Figure 2 shown. Figure 2 If the grayscale value of the central pixel in Figure 2 is 0, and the grayscale values of the eight-neighborhood pixels are 1, 2, 3, 4, 5, 6, 7, and 8 respectively, then

[0027] in

[0028] the texture trend pairs of the central pixel are (0, 1), (0, 2), (0, 3), (0, 4), (0, 5), (0, 6), (0, 7), and (0, 8).

[0029] So far, the texture trend pair set of each pixel in the rendered image is obtained, which is used to characterize the feature of the structural change between the pixel and its surrounding pixels, so as to facilitate the recognition of the similar structural direction of the central pixel.

[0030] Furthermore, different structures have differences in grayscale values in the rendered image. Therefore, the direction with the largest difference in grayscale values is the direction of the transformation of different structures. Thus, for each central pixel, calculate the absolute value of the difference in grayscale values between the central pixel and its eight-neighborhood pixels, and select the neighborhood pixel with the largest absolute value and the central pixel to form a trend distribution pair, which is used to characterize the feature of the structural change of the central pixel. ; in the formula, represents the structural direction distribution degree between the trend distribution pair of pixel u and the j-th texture trend pair of the i-th pixel in its eight-neighborhood, , respectively represent the first and second elements in the tendency distribution pair of pixel u; , represent the first and second elements in the j-th texture tendency pair of the i-th pixel in the eight-neighborhood of pixel u; || represents the absolute value.

[0031] Further, according to the structural direction distribution degree between the tendency distribution pair of each pixel and all texture tendency pairs of each pixel in its eight-neighborhood, calculate the structural distribution feature degree between each pixel and each pixel in its eight-neighborhood. In this embodiment, the specific calculation relationship is: ; where represents the structural distribution feature degree between pixel u and the i-th pixel in its eight-neighborhood; represents the set composed of the structural direction distribution degrees between the tendency distribution pair of pixel u and all texture tendency pairs of the i-th pixel in its eight-neighborhood; min() represents the minimum value function.

[0032] When there is no noise in the rendered grayscale image, the pixels in the eight-neighborhood of the central pixel are of the same structural substance, then there are other continuous pixels with the same structure as the central pixel in the eight-neighborhood. Therefore, in the case of no noise, the smaller the difference between the tendency distribution pair of the central pixel and the texture tendency pairs of the remaining pixels, that is the smaller, making the minimum value also smaller. Therefore, the structural distribution feature degree between the eight-neighborhood pixels and the central pixel is smaller, which is used to analyze the direction with the most similar structure between the neighborhood pixels and the central pixel to evaluate whether the neighborhood pixel and the central pixel are of the same texture structure.

[0033] In addition, when the difference between the tendency distribution pair of the central pixel and the texture tendency pairs of the eight-neighborhood pixels is large, it indicates that the central pixel is at the junction of two different structural substances. Since the junction of different substances is not necessarily linear and may also be arc-shaped, therefore, the structural distribution feature degree is calculated between the texture tendencies in multiple directions of the pixel and the central pixel. When the local structural distribution of the pixels in the eight-neighborhood is similar to the local structural distribution of the central pixel, that is, there is a smaller difference between the texture tendency pair in the direction where the neighborhood pixel is located and the tendency distribution pair of the central pixel, making the structural distribution feature degree between the central pixel and the neighborhood pixel smaller, so as to better distinguish the structural distribution features between the pixels in the rendered grayscale image.

[0034] So far, according to the above process of this embodiment, the structural distribution feature degree between each pixel and each pixel in its eight-neighborhood can be obtained. Further, the structural distribution feature degrees between each pixel and all pixels in the eight-neighborhood are combined to form the structural distribution feature sequence of each pixel.

[0035] By performing the above steps, the structural distribution feature sequence of each pixel of the rendered grayscale image is obtained. Thus, the adjacent structure difference degree of each pixel of the rendered grayscale image is calculated. Taking any pixel as an example: when the central pixel is not at the boundary of the structure, the texture feature difference between the central pixel and the pixels of the same structure within its eight-neighborhood is relatively small; when the central pixel is at the boundary of the structure, the texture feature difference between the central pixel and the pixels within its eight-neighborhood is relatively large. Therefore, the structural distribution feature sequences of all pixels are used as the input of the K-means clustering algorithm, where the value of the necessary parameter clustering number K is 3, and the difference degree of the structural distribution features between pixels is used as the clustering metric distance to classify the pixels, and the output is 3 clustering clusters. Among them, the calculation of the K-means clustering algorithm is a well-known technology, and the specific calculation process will not be elaborated here.

[0036] Furthermore, for each pixel's clustering cluster, the more the number of pixels in the clustering cluster, the more pixels are similar to the local structural distribution features of that pixel. If this pixel is a structural edge point, it indicates that the complexity of this pixel is higher. To ensure the rendering effect, the filtering degree at the position of this pixel is smaller. For each pixel, obtain the texture trend pair set composed of the texture trends corresponding to all pixels in the clustering cluster where the pixel is located, which is used to represent the set of pixels similar to the texture features of this pixel, denoted as . According to the difference situation of the texture trend pairs of all pixels within each pixel's clustering cluster, calculate the adjacent structure difference degree of each pixel. In this embodiment, the specific calculation relationship is: ; in the formula, represents the adjacent structure difference degree of pixel u; N represents the number of texture trend pairs in the texture trend set of pixel u; , respectively represent the first and second elements of the k-th texture trend pair in the texture trend set of pixel u.

[0037] When the pixel is at the edge of the structure, the difference between this pixel and other structure pixels is greater, that is, the absolute value of the difference in gray values between pixels is greater. During the process of filtering the scene image, the more details of this pixel should be retained. Therefore, the adjacent structure difference degree calculated by this pixel based on the texture feature difference in the eight-neighborhood is greater to ensure that more detail information is retained during the filtering rendering of the virtual scene image; conversely, the central pixel is inside the structure, that is, there is less detail information in the local neighborhood of the pixel, and the texture structure is relatively uniform, and lower detail attributes can be retained, thereby reducing the time and energy consumption during the rendering process.

[0038] So far, according to the above process of this embodiment, the proximity structure difference degree of each pixel in the rendered grayscale image can be obtained, which is used to characterize the structural change situation within the neighborhood of each pixel.

[0039] Step 4: Analyze the dispersion degree of the grayscale values of the pixels within the local neighborhood of each pixel, and combine with the proximity structure difference degree of each pixel to obtain the rendering blurriness of each pixel, perform a linear mapping on the rendering blurriness of all pixels to obtain the corresponding mapping value for each pixel, and use the guided image filtering algorithm to filter the rendered image and perform rendering processing on the filtered image.

[0040] For the denoising and enhancement of virtual scene images, since during the rendering of virtual scenes, the higher the detail attributes between different structures in the virtual scene image, the better the rendering effect of the virtual scene after rendering. However, high-detail attribute images require a longer rendering time, resulting in a poor real-time effect for virtual reality technology and bringing higher energy consumption. Therefore, it is necessary to perform denoising and enhancement on the image, retain the detail attributes at the junctions of different structures, and simplify the detail attributes within the same structure, so that the virtual reality technology can render the virtual scene image faster during the rendering process.

[0041] It can be understood that the proximity structure difference degree of each pixel represents the degree of structural change at each position of the rendered grayscale image. The greater the proximity structure difference degree of each pixel, the greater the possibility that the pixel at the corresponding position in the rendered grayscale image is at the structural edge. For this part, greater detail attributes should be retained, and after rendering using virtual reality technology at this position, there are obvious layers between different structures.

[0042] Based on the above analysis, by the grayscale dispersion degree within the local neighborhood of each pixel in the rendered grayscale image and the proximity structure difference degree of each pixel, calculate the rendering blurriness of each pixel in the rendered grayscale image. In this embodiment, the specific calculation relationship is: ; In the formula, represents the rendering blurriness of pixel u; represents the standard deviation of the grayscale values of all pixels within the eight-neighborhood of pixel u; represents the proximity structure difference degree of pixel u; Norm() is a normalization function; represents a constant to avoid a zero denominator, with a value range from 0 to 0.1. In this embodiment, to prevent the denominator from being zero, the value is taken as 0.001.

[0043] When there are obvious differences in the local distribution of pixels in the rendered grayscale image, it indicates that this position corresponds to different texture structures. Therefore, the standard deviation of the grayscale values of all pixels within the eight-neighborhood of the pixel The larger it is, the more detailed attributes should be retained at the position of the pixel point in the rendered grayscale image, so as to ensure better rendering effects through virtual reality technology. Therefore, the rendering blur degree of this pixel point is smaller, thereby enhancing the rendering effect of virtual reality technology on virtual scenes.

[0044] Furthermore, in this embodiment, the rendering blur degrees of each pixel point of the rendered image are combined to form a rendering blur image. In image processing, when the local detailed attributes are larger, it means that the local changes and features in the image are more abundant and complex. In this case, in order to better retain these details and features, the truncation value ε should be smaller. Thus, based on the above analysis, all the values in the rendering blur image are linearly mapped to , and the corresponding mapped values of each pixel point are obtained. The mapped values of all pixel points form a mapped image. Among them, the calculation of linear mapping is a well-known technology, and the specific calculation process will not be elaborated here.

[0045] Replace the truncation value in the guided image filtering with the mapped image obtained in the above process to improve the guided image filtering, that is, use the mapped value of each pixel point as the truncation value in the guided filtering process to perform filtering processing on each pixel point. Further, use the rendered image as the input of the improved guided image filtering. Among them, the value of the necessary parameter filtering window r takes 13 in this embodiment, and the output is the filtered and enhanced image.

[0046] Furthermore, transmit the filtered and enhanced image to the real-time rendering engine. Combining rasterization and ray tracing technologies, utilize the parallel computing ability of the GPU to achieve high-speed rendering of multiple frames per second. Real-time rendering engines such as Unreal Engine and Unity. The rendering process is a commonly used technology in the field of video rendering, and the specific operation steps will not be elaborated here.

[0047] It can be understood that: the above sequence of embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above description of specific embodiments of this specification is given. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0048] Each embodiment in this specification is described in a progressive manner. The same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0049] The above content is only an implementation manner of this application and is not intended to limit the scope of this application. Any equivalent structural or equivalent process transformation made by using the content of the specification and drawings of this application, or directly or indirectly applied to other related technical fields, shall similarly be included within the protection scope of this application.

Claims

1. A method for enhancing the realism of a digital twin model based on a rendering engine, characterized in that Including the following steps: Obtain a rendered image for virtual scene rendering, and perform grayscale processing to obtain a rendered grayscale image; Construct a texture trend pair for each pixel point based on the distribution characteristics of the gray values of the pixel points in the local neighborhood of each pixel point in the rendered grayscale image, and select a trend distribution pair for each pixel point according to the gray difference between each pixel point and the pixel points in the local neighborhood; Analyze the differences between the trend distribution pairs of each pixel point and the texture trend pairs of each pixel point in the local neighborhood to obtain the structural distribution feature degree between each pixel point and each pixel point in the local neighborhood, and perform clustering division on all pixel points based on the structural distribution feature degree. Obtain the adjacent structure difference degree of each pixel point through the difference situation of the texture trend pairs of all pixel points in the cluster where each pixel point is located after clustering division; Analyze the dispersion degree of the gray values of the pixel points in the local neighborhood of each pixel point, and combine the adjacent structure difference degree of each pixel point to obtain the rendering blur degree of each pixel point, and perform linear mapping on the rendering blur degrees of all pixel points to obtain the corresponding mapping values for each pixel point. Use the guided image filtering algorithm to filter the rendered image, and perform rendering processing on the filtered image.

2. The method for enhancing the realism of a digital twin model based on a rendering engine according to claim 1, characterized in that The construction of the texture trend pair for each pixel point includes: for each pixel point, respectively form the texture trend pair of each pixel point with the gray values of its eight neighborhood pixel points.

3. The method for enhancing the realism of a digital twin model based on a rendering engine according to claim 2, wherein The determination of the trend distribution pair for each pixel point includes: calculate the absolute value of the gray difference between each pixel point and its eight neighborhood pixel points, and select the texture trend pair corresponding to the neighborhood pixel point with the largest absolute value as the trend distribution pair for each pixel point.

4. The method for enhancing the realism of a digital twin model based on a rendering engine according to claim 2, wherein, The corresponding calculation method for the structural distribution feature degree between each pixel point and each pixel point in the local neighborhood is: Calculate the differences between the trend distribution pairs of each pixel point and the texture trend pairs of the pixel points in the eight neighborhoods to obtain the structural direction distribution degree between the trend distribution pair of each pixel point and each texture trend pair of the pixel points in the eight neighborhoods; Then the expression for the structural distribution feature degree between each pixel point and each pixel point in the local neighborhood is as follows: ; represents the structural distribution feature degree between pixel point u and the i-th pixel point in its eight-neighborhood; represents the set composed of the structural direction distribution degrees between all texture trend pairs of the trend distribution pair of pixel point u and the i-th pixel point in its eight-neighborhood; min() represents the minimum value function.

5. The method for enhancing the realism of a digital twin model based on a rendering engine according to claim 4, wherein The corresponding calculation formula for the structural direction distribution degree between the trend distribution pairs of each pixel and each texture trend pair of the pixels in the eight-neighborhood is as follows: ; In the formula, represents the structural direction distribution degree between the trend distribution pair of pixel u and the j-th texture trend pair of the i-th pixel in its eight-neighborhood, and respectively represent the first and second elements in the trend distribution pair of pixel u; and represent the first and second elements in the j-th texture trend pair of the i-th pixel in the eight-neighborhood of pixel u; || represents the absolute value.

6. The method for enhancing the realism of a digital twin model based on a rendering engine according to claim 1, wherein During the process of clustering division of all pixel points, it further includes: forming the structural distribution feature sequence of each pixel point with the structural distribution feature degrees between each pixel point and all pixel points in the eight neighborhoods, using the structural distribution feature sequences of all pixel points as the input of the clustering algorithm, and using the difference between the structural distribution feature degrees between any two pixel points as the metric distance in the clustering process to perform clustering division on the pixel points.

7. A method for enhancing the realism of a digital twin model based on a rendering engine according to claim 1, characterized in that, The calculation method for the proximity structure difference degree of each pixel is as follows: ; In the formula, represents the proximity structure difference degree of pixel u; N represents the number of texture trend pairs in the texture trend set of pixel u; and respectively represent the first and second elements of the k-th texture trend pair in the texture trend set of pixel u; Among them, the texture trend set is the set composed of the texture trend pairs of all pixel points in the cluster where the pixel point is located.

8. A method for enhancing the realism of a digital twin model based on a rendering engine according to claim 1, characterized in that The corresponding calculation method for the rendering blur degree of each pixel is as follows: ; In the formula, represents the rendering blur degree of pixel u; represents the standard deviation of the gray values of all pixels within the eight-neighborhood of pixel u; represents the difference degree of the adjacent structure of pixel u; Norm() is a normalization function; represents a constant to avoid a zero denominator.

9. A method for enhancing the realism of a digital twin model based on a rendering engine according to claim 1, characterized in that The step of using the guided image filtering algorithm to filter the rendered image further includes: using the mapping value corresponding to each pixel point as the truncation value in the process of filtering the rendered image by the guided filtering algorithm to perform filtering processing on each pixel point in the rendered image.

10. A method for enhancing the realism of a digital twin model based on a rendering engine according to claim 1, characterized in that, The grayscale processing uses the weighted average method.

Images