CN116228984A - A Volumetric Cloud Modeling and Rendering Method Based on Meteorological Data - Google Patents

Abstract

The present invention proposes a volumetric cloud modeling and rendering method based on meteorological data, which includes five data fields of cloud base height, cloud top height, coverage, cloud base type, and cloud top type. First, the cloud data structure is constructed basically, The depth value of the volumetric cloud, the ray stepping algorithm, the rendering scene is mixed and the density of the cloud in different regions is completely controlled by the noise parameter, and the Perlin noise and Worley noise of different frequencies are superimposed and saved as a three-dimensional texture for volumetric For cloud modeling, the shape noise texture uses fractal Perlin-Worley noise to generate the basic shape of the cloud; the detail noise texture uses fractal Perlin noise to erode the edge of the cloud shape to add details, realizing the static Very realistic volumetric clouds, and the cloud-penetrating interaction with volumetric clouds in the moving state, this technical solution can not only reflect the cloud attitude driven by real data, but also use a lower rendering cost to obtain a more realistic volume cloud effect.

Description

A Volumetric Cloud Modeling and Rendering Method Based on Meteorological Data

technical field

The invention relates to the technical field of volumetric cloud modeling and rendering, in particular to a volumetric cloud modeling and rendering method based on meteorological data.

Background technique

To model the volumetric cloud is to construct the density field of the volumetric cloud and calculate the density value of the cloud at any point in the space. The current mainstream approach is generally to use 3D textures to store cloud density information. 3D textures can directly store cloud density values, distance field information, and baked lighting information. There are various specific methods, depending on the needs. depends.

This article mainly introduces the real-time sky volumetric cloud modeling method. At present, the mainstream method of PC/host end is to carry out procedural modeling based on noise texture, and then cooperate with some art resources to make more abundant forms and performance effects. There are also many ways to use noise textures and art resources. For example, the methods shared by Horizon, Red Dead Redemption 2, and Frost are all worthy of reference. In UE4, Material Graph is even used directly, and the modeling method is completely open. For art/TA, although this method is very flexible, it is not easy to control and optimize performance. After struggling for a long time, this method has not been adopted yet.

Among them, the "a three-dimensional cloud scene modeling and visualization method based on ground-based cloud image" disclosed by the application number "CN202111589721.7" is also an increasingly mature technology. , calculate the cloud base height and cloud thickness as the parameters of three-dimensional cumulus modeling. Determine the position information of all voxel center points that constitute the cumulus model according to the above two parameters, and pass it to the GPU as a vertex array, and use it in the geometric coloring of the GPU The shader stage draws all cumulus voxels according to the vertex array, and finally enters the fragment shader stage of the GPU to determine the density value of the cloud according to the distance between each voxel and the center point of the cloud, and calculate the color of the voxel according to the density value to realize the three-dimensional cloud Integration of scene modeling and visualization. The present invention takes voxel as the smallest geometric unit, builds a three-dimensional cumulus model based on GPU and realizes visualization in a three-dimensional scene, improves the efficiency of three-dimensional cumulus modeling, and satisfies the requirements of three-dimensional cumulus modeling in three-dimensional geographical scenes. Need for smooth roaming”, but this kind of agricultural production traceability system still has the following defects in the actual use process:

The existing technical methods cannot build the basis of the volumetric cloud, so that the density field of the volumetric cloud cannot be constructed, the rendering accuracy is limited, and the authenticity of the volumetric cloud is obtained poorly.

Contents of the invention

The purpose of the present invention is to provide a volumetric cloud modeling and rendering method based on meteorological data, so as to solve the problem that the above-mentioned background technology cannot construct the basis for the volumetric cloud, so that the density field of the volumetric cloud cannot be constructed, and the accuracy of rendering cannot be solved. Limited, the problem that the authenticity of the obtained volumetric cloud is poor.

To achieve the above object, the present invention provides the following technical solutions: a volumetric cloud modeling and rendering method based on meteorological data, constructing a volumetric cloud data structure, including cloud base height, cloud top height, coverage, cloud base type, cloud top type These five data fields; firstly, the data structure of the cloud is basically constructed, the depth value of the volumetric cloud, the light stepping algorithm is used, the rendering scene is mixed, and the density of the cloud in different regions is completely controlled by the noise parameter;

The superimposed Perlin noise and Worley noise of different frequencies are saved as three-dimensional textures for modeling volumetric clouds;

The shape noise texture uses fractal Perlin-Worley noise to generate the basic shape of the cloud; the detail noise texture uses fractal Perlin noise to erode the shape edge of the cloud to add details;

The volumetric clouds are produced from the analysis of shadows.

Preferably, the basic cloud profile is constructed according to the data structure of the volumetric cloud, and then three-dimensional noise is used to erode the above cloud profile to add details to the cloud layer to obtain the density field of the volumetric cloud.

Preferably, the acquisition of the depth value of each pixel of the scene to be displayed and the depth value of the volumetric cloud compares and determines whether each pixel is blocked by clouds, and if the pixel is not blocked by clouds, the original color of the pixel is directly displayed.

Preferably, the ray stepping algorithm is used to obtain the intersection point of the ray emitted by the virtual camera and the surface of the cloud layer as the starting point. If the viewpoint is located in the cloud, a structured sampling method is adopted, and the viewpoint position is used as the starting point. Step along the direction of the ray, and obtain the concentration of the sampling point in the density field.

Preferably, the illumination model is solved according to the self-occlusion shadow and the cloud concentration of the sampling point to obtain the final transparency and color of the volumetric cloud, and mix it with the scene to be rendered.

Preferably, the density of clouds according to different regions is completely controlled by noise parameters, if the artist needs to have different types of clouds in the scene at the same time, or needs to be able to freely control the density of clouds.

Preferably, for cumulonimbus clouds with complex shapes, it is necessary to combine Cloud Lut to transition between Cloud Type 0.5 to 1.0 (this can of course be done), that is, to give noise different coverage and erosion degrees according to height, High cloud coverage is low and noise erosion is high, and low cloud coverage is high and noise erosion is low.

Preferably, there are mainly two methods for the self-shadowing of the volume cloud:

The first method is more common. The method is to do RayMarching (PC/host side) for each sampling point in the direction of the light source to calculate the transmittance. The effect is better when the number of sampling steps is sufficient, but obviously the overhead will be high. bigger;

The second method is volumetric shadow mapping, which uses UE4's Beer's Shadow Map, and the more complicated Transmittance Fucntion Mapping method (currently only seen in Final Fantasy). The principle of the Transmittance Fucntion Mapping method is to use a A series of orthogonal basis functions to approximate the transmittance function.

Compared with prior art, the beneficial effect of the present invention is:

When performing volumetric cloud modeling, the basic shape of the cloud layer is constructed according to the cloud coverage data and cloud type data, and a variety of noises of different frequencies are superimposed to add details, thereby constructing the density field of the volumetric cloud. When rendering, use the ray stepping method to calculate the cumulative concentration and light shading of a single pixel, solve the lighting model to get the final color and transparency of the volumetric cloud, and then mix it with the scene. Different sampling methods are used for different usage scenarios to achieve In addition to the realistic volumetric cloud when the viewpoint is still, and the cloud-penetrating interaction with the volumetric cloud in the moving state, this technical solution can not only reflect the cloud attitude driven by real data, but also use a lower rendering cost to obtain realistic Higher volumetric cloud effect.

Description of drawings

Fig. 1 is a flow chart of the volumetric cloud construction and rendering method of the present invention.

Figure 2 is a height distribution diagram of the present invention.

Fig. 3 is the low frequency Perlin Worley noise of the present invention and the Worley noise figure that the frequency increases successively.

FIG. 4 is a detailed view of volumetric cloud augmented with 3D noise in the present invention.

Fig. 5 is a schematic flow chart of structured sampling in the present invention.

Detailed ways

The technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Apparently, the described embodiments are part of the embodiments of the present invention, but not all of them.

The components of the embodiments of the invention generally described and shown in the figures herein may be arranged and designed in a variety of different configurations. Accordingly, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the invention.

Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" etc. The indicated orientation or positional relationship is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, or in a specific orientation. construction and operation, therefore, should not be construed as limiting the invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and should not be construed as indicating or implying relative importance.

Referring to Fig. 1-Fig. 5, the present invention provides a kind of technical scheme: structure volumetric cloud data structure, wherein comprise these five data fields of cloud base height, cloud top height, coverage rate, cloud base type, cloud top type; First the data of cloud The basic structure construction, the depth value of the volume cloud, the use of light stepping algorithm, the rendering scene is mixed, and the density of the cloud in different regions is completely controlled by the noise parameter; the Perlin noise and Worley noise of different frequencies are superimposed and saved as a three-dimensional texture. Used to model volumetric clouds; the shape noise texture uses fractal Perlin-Worley noise to generate the basic shape of the cloud; the detail noise texture uses fractal Perlin noise to erode the edge of the cloud shape to add details;

The analysis and production of the volumetric cloud self-shadow, the basic cloud outline is constructed according to the data structure of the volumetric cloud, and then the three-dimensional noise is used to erode the above cloud outline, add details to the cloud layer, and obtain the density field of the volumetric cloud.

Obtain the depth value of each pixel of the scene to be displayed and the depth value of the volumetric cloud, compare and determine whether each pixel is blocked by clouds, if the pixel is not blocked by clouds, directly display the original color of the pixel, and use the ray stepping algorithm to obtain The intersection point of the ray emitted by the virtual camera and the surface of the cloud layer is used as the starting point. If the viewpoint is in the cloud, the structured sampling method is adopted, and the viewpoint position is used as the starting point. From the starting point, the point along the direction of the ray is taken to obtain samples in the density field. point concentration.

Solve the lighting model according to the self-occlusion shadow and the cloud concentration of the sampling point, get the final transparency and color of the volumetric cloud, and mix it with the scene to be rendered. The density of the cloud in different areas is completely controlled by the noise parameter. If the art needs the scene There are different types of clouds at the same time, or you need to be able to freely control the density of clouds, you need to use Cloud Map, Cloud Map can be drawn directly by artists, but because this method cannot observe the effect in real time, it is not convenient, So we made a brush tool to facilitate the drawing of artists, similar to UE4's Blueprint PaintedClouds.

For cumulonimbus clouds with complex shapes, it is necessary to combine Cloud Lut to transition between Cloud Type 0.5 to 1.0 (of course this can also be done), that is, to give noise different coverage and erosion degrees according to height, and cloud coverage at high places Low cloud coverage and high noise erosion, low cloud coverage and low noise erosion;

The Erosion stored in the G channel in CloudLut is not only used for detail noise, but also affects the calculation result of shape noise. For details, please refer to the implementation in the Unity HDRP source code.

There are two main methods for volumetric cloud self-shadowing: the first method is more common, and the method is to do RayMarching (PC/host side) for each sampling point in the direction of the light source to calculate the transmittance. In the case of enough sampling steps The effect is also better, but obviously the overhead will be relatively high; the second method is volumetric shadow mapping, using UE4's Beer'sShadow Map, and a more complicated Transmittance Fucntion Mapping method (currently only seen similar methods used in Final Fantasy ), the principle of the Transmittance Fucntion Mapping method is to use a series of orthogonal basis functions to approximate the transmittance function. The selection of basis functions can be done by Fourier function, DCT function, Haar wavelet function, etc. Compared with Beer's Shadow Map, TFM Such methods can approximate the transmittance curve more accurately in some cases, resulting in better rendering of volumetric clouds.

In the volumetric cloud modeling process, the volumetric cloud weather map is obtained according to the real meteorological data, and a three-channel two-dimensional texture is made. The r channel stores the coverage information of the volumetric cloud, the g channel stores the cloud base type information, and the b channel stores the cloud top type information . The volumetric cloud coverage represents the occurrence probability p _{h horizontal} of the volumetric cloud on the horizontal layer. The value range of cloud layer type data is 0.0～1.0. According to the type of cloud layer, the volume cloud occurrence probability p vertical in the vertical direction is obtained from the vertical distribution diagram. According to these two _{probabilities} , the volume cloud existence probability field p is calculated according to the following formula _profile , this probability field depicts the approximate outline of the volumetric cloud.

p _profile =p _{h horizontal} *P _vertical

Use the Fractal Brownian Motion method to superimpose multiple noises with sequentially increasing frequencies and sequentially decreasing amplitudes, and generate two 3D noise textures offline. The calculation formula of FBM is as follows:

Among them, n is the number of superimposed textures, amplitude represents the amplitude of the noise superimposed each time, frequency represents the frequency of the superimposed noise, and the noise() function represents the noise function used. Using the fbm method can also obtain the dynamic effect of the volumetric cloud, only need to perform domain distortion when reading the noise texture, and use the following formula to calculate the noise value after domain distortion:

Density(p)=fbm(p+fbm(p+fbm(p)))

Procedurally generated noise is pseudo-random and is usually pre-generated and saved as an image. Volumetric cloud modeling requires two 3D noise textures. The size of the first texture is 128*128*128*4, the R channel stores low-frequency PelinWorley noise, and the GBA channel stores Worley noise with increasing frequency.

Low-frequency Perlin Worley noise and Worley noise with increasing frequency, the resolution is 128 ³ , the size of the second 3D noise texture is 32*32*32*3, and the three channels store Worley noise with higher frequency respectively, which is used for volume Clouds add more detail, high-frequency Worley noise, with a resolution of 32 ³ ;

Finally, the density field of the volumetric cloud is calculated by

Where noise represents the passed or distorted noise value;

Rendering process: selection of sampling points, using the ray stepping method to generate a ray pointing to the volumetric cloud from the virtual camera position, starting from the camera position, advancing with a specific step, calculating the position of each sampling point, and sampling the cloud at that point Density, accumulate cloud density and calculate the brightness information of each sampling point, and update the color brightness of the volumetric cloud segment pointed to by this ray.

Pixels that are farther away from the camera have less influence on the final rendering state of the volumetric cloud. Therefore, in the volumetric cloud ray stepping method, the method of variable step size sampling is adopted.

When the virtual camera position is on the ground, calculate the intersection point of the step ray and the bottom of the cloud layer, and use this intersection point as the step starting point for variable step size sampling. Stepping light with the same step length for each pixel will lead to obvious banding in the volume cloud. In order to prevent this from happening and speed up rendering efficiency, blue noise is used to add a jitter value to the step size ;

When the virtual camera position is in the cloud, the camera position is used as the starting point of the ray step, and the step size is changed to advance sampling. When the virtual camera is in the cloud, it usually keeps moving, and the change of the viewpoint will lead to a rapid change of the sampling point. This results in temporal aliasing of volumetric clouds. Using blue noise jitter sampling can alleviate this kind of aliasing, but the optimization effect is limited. Using structured sampling method can solve this problem, and calculate the initial displacement of the ray step according to the position of the viewpoint, so that when the viewpoint moves, only the closest and the most Distant sampling points will change. The nearby sampling points are relatively dense, and their changes have negligible influence on the volumetric cloud results. The farthest sampling point itself has a relatively small weight in the density accumulation of a ray step, and its influence on the result is almost negligible.

If the cloud density value of the sampling point is greater than 0, start from the sampling point and sample step by step toward the light source, accumulate the cloud density of the sampling point in the direction of the light source, and solve the illumination model to calculate the brightness of the sampling point.

Lighting model solution:

The lighting model of the volume cloud includes the light and shade effects caused by multiple scattering and refraction in the cloud layer, the bright edge effect when facing the sun directly, and the dark edge effect when facing away from the sun.

Aiming at the phenomenon of multiple scattering and refraction in clouds, Beer's law is used to describe the relationship between volumetric cloud transparency and optical thickness.

cloud_transmittnnce=e ^-Cloud_depth

Among them, Cloud_depth represents the accumulated density of light stepping to the current sampling point, and Density(p) obtains the density value corresponding to the sampling point at p.

For the bright edge effect when facing the direction of the sun, it is caused by the scattering of sunlight through the clouds. The phase function describes the relationship between the scattering intensity of light at various angles and the direction of incident light, which can be used to simulate the scattering of light in clouds. MIE scattering can describe this scattering relationship more accurately, but the calculation cost is high, so the HG phase function is usually used to approximate MIE scattering in cloud rendering. In order to make up for the large approximation error of the HG phase function in backscattering, the TTHG phase function is used instead, and the parameters in TTHG are optimized using the particle swarm optimization algorithm.

phase _TTHG (θ, g _α , g _β , α) = α·phase _HG (θ, g _α )+(1-α)·phase _HG (θ, g _β )

Where g _α and g _β represent the asymmetry factors of forward scattering and back scattering respectively, and α is the weight;

When observing the clouds facing away from the sun, you will see that the clouds have a dark edge effect, which cannot be simulated by the above two methods alone. When viewing volumetric clouds facing away from the sun, use the Beer's-Powder function instead of the Beer function.

E＝2.0·e ^-Cloud_depth ·(1.0-e ^{-2·Cloud_depth} )

To reduce the difference between the result and the original normalized function, the Beer's-Powder function multiplies the result by 2.

The final lighting model can be expressed by the following formula:

E＝2.0·e ^-Cloud_deoth ·(1.0-e ^{-2·Cloud_depth} )

According to the depth information of the volumetric cloud outline and the depth information of the scene to be rendered, find the pixels blocked by the volumetric cloud. For these pixels, perform the above-mentioned ray stepping respectively, calculate the color and transparency of the volumetric cloud on each pixel, and finally Rendered scenes are blended together.

The content that is not described in detail in this description belongs to the prior art known to those skilled in the art. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still implement the foregoing Modifications to the technical solutions recorded in the examples, or equivalent replacements for some of the technical features, within the spirit and principles of the present invention, any modifications, equivalent replacements, improvements, etc., shall be included in the scope of protection of the present invention within.

Claims (8)

1. A volumetric cloud modeling and rendering method based on meteorological data, constructing a volumetric cloud data structure, which includes five data fields of cloud base height, cloud top height, coverage, cloud base type, and cloud top type; first, cloud data The basic structure construction, the depth value of the volume cloud, the use of light stepping algorithm, the rendering scene is mixed, and the density of the cloud in different regions is completely controlled by the noise parameter; The superimposed Perlin noise and Worley noise of different frequencies are saved as three-dimensional textures for modeling volumetric clouds; The shape noise texture uses fractal Perlin-Worley noise to generate the basic shape of the cloud; the detail noise texture uses fractal Perlin noise to erode the shape edge of the cloud to add details; The volumetric clouds are produced from the analysis of shadows. 2. A kind of volumetric cloud modeling based on meteorological data, rendering method according to claim 1, it is characterized in that: described according to the data structure of volumetric cloud construction basic cloud layer profile, use three-dimensional noise to erode above cloud layer again Silhouettes, add detail to clouds, get density fields for volumetric clouds. 3. A volumetric cloud modeling and rendering method based on meteorological data according to claim 1, characterized in that: the depth value of each pixel of the scene to be displayed and the depth value of the volumetric cloud are acquired, and each pixel is determined by comparison Whether it is blocked by clouds, if the pixel is not blocked by clouds, the original color of the pixel will be displayed directly. 4. a kind of volumetric cloud modeling based on meteorological data, rendering method according to claim 1, is characterized in that: described adopting ray-stepping algorithm, obtains the intersection point that virtual camera sends out ray and cloud layer surface as starting point, if If the viewpoint is located in the cloud, the structured sampling method is adopted. The viewpoint position is used as the starting point, and the concentration of the sampling point is obtained in the density field by stepping along the ray direction from the starting point. For each sampling point with a density greater than 0, the ray stepping algorithm is performed again in the direction of light to calculate the volumetric cloud self-occlusion shadow. 5. a kind of volumetric cloud modeling based on meteorological data, rendering method according to claim 1, is characterized in that: described according to the cloud density solution illumination model of self-occlusion shadow and sampling point, obtains the final transparency of volumetric cloud and color and blend it with the scene to be rendered. 6. A kind of volumetric cloud modeling and rendering method based on meteorological data according to claim 1, characterized in that: the density of the clouds according to different regions is completely controlled by noise parameters, if the art requires that the scene simultaneously There are different types of clouds, or you need to be able to freely control the density of clouds, you need to use CloudMap, CloudMap can be drawn directly by art, but because this method cannot observe the effect in real time, it is not convenient, so we made A brush tool, which is convenient for artists to draw, similar to UE4's BlueprintPaintedClouds. 7. A volumetric cloud modeling and rendering method based on meteorological data according to claim 1, characterized in that: for cumulonimbus clouds with relatively complex shapes, it is necessary to combine CloudLut between 0.5 and 1.0 of CloudType Transition (of course this can also be done), that is, give different coverage and erosion to the noise according to the height, the cloud coverage at high places is low and the noise erosion is high, and the cloud coverage at low places is high and the noise erosion is low; The Erosion stored in the G channel in CloudLut is not only used for detail noise, but also affects the calculation result of shape noise. For details, please refer to the implementation in the UnityHDRP source code. 8. A kind of volumetric cloud modeling and rendering method based on meteorological data according to claim 1, characterized in that: the way of self-shadowing of the volumetric cloud mainly contains two kinds: The first method is more common. The method is to do RayMarching (PC/host side) for each sampling point in the direction of the light source to calculate the transmittance. The effect is better when the number of sampling steps is sufficient, but obviously the overhead will be high. bigger; The second method is volume shadow mapping, which uses UE4's Beer'sShadowMap, and the more complicated TransmittanceFucntionMapping method (only seen in Final Fantasy). The principle of the TransmittanceFucntionMapping method is to use a series of orthogonal bases function to approximate the transmittance function, the selection of basis functions can be done by Fourier function, DCT function, Haar wavelet function, etc. Compared with Beer'sShadowMap, methods such as TFM can approximate the transmittance curve better in some cases. Accurate, resulting in better rendering of volumetric clouds.

Images