Hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation

Abstract

The application provides a hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation, and relates to the technical field of remote sensing data processing, aiming to solve the technical problems of high cost of hyperspectral remote sensing image data acquisition, few data sources and low resolution in the prior art. The method comprises the following steps: determining three-dimensional scene data according to user requirements; inputting the three-dimensional scene data into a pre-constructed hyperspectral imaging simulation model and performing the following operations: based on a global coordinate system, starting from the origin, emitting randomly distributed light rays to the three-dimensional scene object through a virtual imaging plane, the three-dimensional scene object is used to simulate the optical response of the real ground object structure after the light rays are incident; determining the intersection point of each light ray and the three-dimensional scene object; according to the geometric information and material information corresponding to the intersection point, combining the atmospheric radiation parameters, the radiation information of the light ray is calculated and the radiation data at the entrance pupil of the remote sensor is obtained; generating a hyperspectral remote sensing picture according to the radiation data at the entrance pupil.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of remote sensing data processing, and more particularly, to a hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation. BACKGROUND

[0002] Hyperspectral remote sensing is a technology for obtaining ground object reflection or radiation information in continuous narrow bands, which can record spatial information and continuous spectral information of a target at the same time, so as to realize fine identification and quantitative analysis of ground objects. Hyperspectral remote sensing technology plays an important role in environmental monitoring, precision agriculture, mineral and geological exploration, ecological research and disaster management research. However, in practical applications, the hyperspectral remote sensing technology is affected by factors such as observation geometry, imaging conditions and atmospheric changes, resulting in limitations in data acquisition capability and consistency. To solve the above problems, the current main method is to use hyperspectral imaging simulation technology to obtain data. This technology combines physical mechanisms and computer simulation to build a controllable digital experiment platform for remote sensing research, and provides an efficient solution for algorithm testing, data expansion and remote sensor performance optimization.

[0003] The current hyperspectral imaging simulation mainly uses two-dimensional image fusion and three-dimensional scene simulation methods. In addition, in terms of hardware and software acceleration, with the development of computers, the latest graphics processing unit (GPU) has special ray tracing support. The model architecture based on the GPU computing core is the development direction of future hyperspectral remote sensing imaging simulation.

[0004] The three-dimensional scene simulation method represented by the ray tracing algorithm requires a large amount of computing resources, which means that a large number of rays need to be used for simulation to achieve the required accuracy. The resulting long running time and resource consumption hinder the practical application of the hyperspectral imaging simulation model. It can be seen that the existing hyperspectral remote sensing image data has technical problems such as high acquisition cost, few data sources and low resolution. SUMMARY

[0005] Therefore, the present application provides a hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation, which aims to solve the technical problems of high acquisition cost, few data sources and low resolution of hyperspectral remote sensing image data in the prior art.

[0006] One aspect of the present application provides a hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation, comprising: determining three-dimensional scene data according to user needs; inputting the three-dimensional scene data into a pre-constructed hyperspectral imaging simulation model, wherein the hyperspectral imaging simulation model is configured with a global coordinate system and a three-dimensional scene object for matching the three-dimensional scene data, and performs the following operations: based on the global coordinate system, starting from the origin, emitting randomly distributed light rays to the three-dimensional scene object through a virtual imaging plane, wherein the virtual imaging plane is used to simulate the imaging surface of a remote sensor, and the three-dimensional scene object is used to simulate the optical response of a real ground object structure after light ray incidence; determining the intersection point of each light ray with the three-dimensional scene object; according to the geometric information and material information corresponding to the intersection point, combining the atmospheric radiation parameters, calculating the radiation data at the entrance pupil of the remote sensor; and generating a hyperspectral remote sensing image according to the radiation data at the entrance pupil.

[0007] According to an embodiment of the present application, the hyperspectral imaging simulation model includes a plurality of triangular facet elements, wherein determining the intersection point of each light ray with the three-dimensional scene object comprises: determining whether the light ray intersects with a triangular facet element; in response to the light ray intersecting with the triangular facet element, determining whether the intersection point is located within the boundary range of the triangular facet element; and in response to the intersection point being located within the boundary range of the triangular facet element, obtaining the geometric information and material information of the intersection point.

[0008] According to an embodiment of the present application, according to the geometric information and material information corresponding to the intersection point, combining the atmospheric radiation parameters, and calculating the radiation data at the entrance pupil of the remote sensor comprises: calculating a first ground surface reflection radiation parameter of the intersection point from direct solar radiation; calculating a second ground surface reflection radiation parameter of the intersection point from scattered solar radiation; calculating a third ground surface reflection radiation parameter of the intersection point from background reflection radiation; and based on the first ground surface reflection radiation parameter, the second ground surface reflection radiation parameter, and the third ground surface reflection radiation parameter, calculating the total radiation data at the entrance pupil of the remote sensor.

[0009] According to an embodiment of the present application, based on the first ground surface reflection radiation parameter, the second ground surface reflection radiation parameter, and the third ground surface reflection radiation parameter, calculating the total radiation data at the entrance pupil of the remote sensor comprises: calculating the sum of the first ground surface reflection radiation parameter and the second ground surface reflection radiation parameter, or calculating the sum of the first ground surface reflection radiation parameter and the third ground surface reflection radiation parameter, to obtain a ground surface zero distance reflection radiance parameter of each light ray; superimposing the ground surface zero distance reflection radiance parameters of all light rays and performing mean value calculation to obtain a ground surface zero distance reflection radiance parameter of each pixel; and based on the ground surface zero distance reflection radiance parameter of each pixel, fusing the path radiation parameter and the cross radiation to obtain the total radiation data at the entrance pupil of the remote sensor.

[0010] According to an embodiment of the present application, the pre-constructed hyperspectral imaging simulation model comprises: establishing a local coordinate system with the optical center of the remote sensor as the origin, and transforming the position and direction of the remote sensor from the local coordinate system to a global coordinate system according to the specified remote sensor observation geometry information; based on the global coordinate system, emitting randomly distributed light rays from the origin to the three-dimensional scene object through a virtual imaging plane; determining the intersection of each light ray with the three-dimensional scene object; according to the geometric information and material information corresponding to the intersection, combining the atmospheric radiation parameters, the radiation data at the entrance pupil of the remote sensor is calculated, wherein the radiation data at the entrance pupil represents the actual received radiation brightness of each pixel; based on the generation path of the radiation data at the entrance pupil, a hyperspectral imaging simulation model is constructed.

[0011] According to an embodiment of the present application, the hyperspectral imaging simulation model is implemented based on a graphics processor, wherein the rendering pipeline corresponding to the graphics application programming interface of the graphics processor comprises: a path tracking renderer, configured to calculate the light transmission path result by using a ray tracing model according to the input three-dimensional scene data, and output and encode the light transmission path result to a texture buffer storage through a pixel shader; a radiance renderer, serially arranged with the path tracking renderer, configured to calculate the radiation data at the entrance pupil of the remote sensor according to the light transmission path result and in combination with the spectral radiation parameters.

[0012] According to an embodiment of the present application, the path tracking renderer is configured to accelerate the ray tracing model by means of Monte Carlo random sampling, wherein the path tracking renderer comprises: a hierarchical bounding box model, configured to accelerate the detection of the intersection features of the light rays and the three-dimensional scene object, wherein the hierarchical bounding box model is configured to search the intersection features of the light rays and the node bounding boxes in turn from the root node according to a depth-first search strategy; an importance sampling module, configured to preferentially select the direction or target with the greatest impact on the result for sampling in the Monte Carlo random sampling process; a low-discrepancy sequence sampling module, configured to generate a uniformly distributed point set in the sample space by mathematical construction to improve the convergence speed of Monte Carlo integration; a simulation accuracy and simulation efficiency balancing module, configured to determine the optimal number of light rays according to the simulation accuracy and simulation efficiency.

[0013] Another aspect of the present application provides a hyperspectral remote sensing image generation device based on three-dimensional scene radiation simulation, comprising: a first determination module configured to determine three-dimensional scene data according to user requirements; an input module configured to input the three-dimensional scene data into a pre-constructed hyperspectral imaging simulation model, wherein the hyperspectral imaging simulation model is configured with a global coordinate system and a three-dimensional scene object for matching the three-dimensional scene data, and performs the following modules: a transmission module configured to transmit randomly distributed light from an origin to the three-dimensional scene object through a virtual imaging plane based on the global coordinate system, wherein the virtual imaging plane is used to simulate an imaging surface of a remote sensor, and the three-dimensional scene object is used to simulate the optical response of a real ground object structure after light incidence; a second determination module configured to determine the intersection of each light and the three-dimensional scene object; a calculation module configured to calculate the radiation data at the entrance pupil of the remote sensor according to the geometric information and material information corresponding to the intersection, combined with atmospheric radiation parameters; and a generation module configured to generate hyperspectral remote sensing picture data according to the radiation data at the entrance pupil.

[0014] Another aspect of the present application provides an electronic device, comprising: one or more processors; a memory storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method as above.

[0015] Another aspect of the present application provides a computer-readable storage medium storing computer-executable instructions that, when executed, perform the method as above.

[0016] Another aspect of the present application provides a computer program product comprising computer-executable instructions that, when executed, perform the method as above.

[0017] Compared with the prior art, the hyperspectral remote sensing image generation device based on three-dimensional scene radiation simulation provided by the embodiments of the present application has at least the following beneficial effects:

[0018] (1) The hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation provided by the embodiment of the present application, based on the ground background and target space-based imaging process derivation, a hyperspectral imaging simulation model is constructed by using the ray tracing algorithm, through the hyperspectral imaging simulation model, starting from the remote sensor, the light is emitted to the three-dimensional scene in reverse from the pixel by pixel, then the propagation path and the radiation information source of each light are tracked, the radiance information of each pixel is calculated, and finally the hyperspectral remote sensing picture data is generated. Since the hyperspectral imaging simulation model is configured with a three-dimensional scene object for simulation, it can quickly simulate the scene under various observation geometric conditions, solar conditions and atmospheric conditions according to user requirements, generate qualified hyperspectral simulation data and images, meet the needs of various fields, and significantly reduce the economic and time cost of hyperspectral remote sensing data acquisition.

[0019] (2) The hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation provided by the embodiment of the present application, the GPU is used to realize the acceleration of the hyperspectral imaging simulation model, that is, the rendering pipeline corresponding to the graphics application programming interface of the graphics processing unit is improved, the path tracking renderer and the radiosity renderer are introduced, the hyperspectral rendering is realized, and the data input and output problem, the quantification problem and the multi-spectral rendering problem of the original rendering pipeline are solved. Since the improved rendering pipeline splits the original ray tracing renderer into two serially executed renderers, the amount of radiation calculation is greatly reduced, and the efficiency of large-scale hyperspectral simulation is improved.

[0020] (3) The hyperspectral remote sensing image generation method and device based on three-dimensional scene radiation simulation provided by the embodiment of the present application, the level bounding box is used to accelerate the intersection of light and scene, the importance sampling and low difference sequence are used to accelerate the convergence of Monte Carlo integration, through the exploration of the relationship between the number of rays and the simulation accuracy, the optimal number of rays is selected for simulation experiment, and the simulation efficiency is greatly improved at a very small accuracy cost. BRIEF DESCRIPTION OF DRAWINGS

[0021] The above and other objects, features and advantages of the present application will become more apparent from the following description of the embodiments of the present application taken with reference to the accompanying drawings, in which:

[0022] Figure 1 A flowchart of the hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation according to the embodiment of the present application is schematically shown.

[0023] Figure 2 A coordinate system diagram of the hyperspectral imaging simulation model according to the embodiment of the present application is schematically shown.

[0024] Figure 3 A display diagram of the sampling point receiving direct solar radiation according to the embodiment of the present application is schematically shown.

[0025] Figure 4 A diagram illustrating a sampling point receiving no direct solar radiation is shown schematically in accordance with an embodiment of the present application.

[0026] Figure 5 A diagram illustrating a sampling point receiving diffuse solar radiation is shown schematically in accordance with an embodiment of the present application.

[0027] Figure 6 A diagram illustrating a sampling point receiving background reflected radiation is shown schematically in accordance with an embodiment of the present application.

[0028] Figure 7 A flowchart illustrating a preliminary improved graphics application programming interface ray-tracing RGB rendering pipeline is shown schematically in accordance with an embodiment of the present application.

[0029] Figure 8 A flowchart illustrating a secondarily improved graphics application programming interface ray-tracing hyperspectral rendering pipeline is shown schematically in accordance with an embodiment of the present application.

[0030] Figure 9 A flowchart and functional module block diagram illustrating a hyperspectral remote sensing imaging simulation software is shown schematically in accordance with an embodiment of the present application.

[0031] Figure 10 A block diagram illustrating a structure of an electronic device adapted to implement a hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation is shown schematically in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0032] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. It is to be understood, however, the embodiments are merely exemplary and the present application is not limited thereto as set forth herein. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present application. However, it would be apparent to those skilled in the art that the present application can be practiced without such specific details. In other instances, well-known structures and

[0033] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used herein, the term "includes" and tautological expressions thereof, means the inclusion of the stated features, steps, operations, and / or components but not to the exclusion of one or more additional features, steps, operations, and / or components.

[0034] All terms used herein, including technical and scientific terms, have the meanings as commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein are to be interpreted as having a meaning that is consistent with the understanding of one of ordinary skill in the art, and not in an overly pedantic or overly formal manner.

[0035] In the case of using expressions similar to "at least one of A, B, and C, etc.", it should generally be interpreted that the expression includes any of one, all of the same, or a combination thereof, unless otherwise defined in the specification. For example, "a system having at least one of A, B, and C" should be interpreted to include a system having A alone, a system having B alone, a system having C alone, a system having both A and B together, a system having both A and C together, a system having both B and C together, and / or a system having all of A, B, and C together, etc.

[0036] In embodiments of the present application, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of the data involved (for example, including but not limited to user personal information) comply with relevant laws and regulations, are used for legal purposes, and do not violate public order and good customs. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain user personal information security and network security.

[0037] Current hyperspectral imaging simulation mainly uses two-dimensional image fusion and three-dimensional scene simulation methods.

[0038] The research of two-dimensional image fusion method mainly focuses on the fusion and spectral reconstruction of existing multi-source data. However, due to the nature of two-dimensional image fusion method, which is a two-dimensional image to image mapping process, there is a limitation in the perspective, which cannot meet the demand of multi-angle and multi-condition hyperspectral imaging simulation.

[0039] Three-dimensional scene simulation method can simulate scenes under arbitrary observation geometry, solar conditions and atmospheric conditions, and has more advantages in the scene of pursuing more quantitative radiation calculation.

[0040] In addition, in terms of hardware and software acceleration, with the development of computers, the latest graphics processing unit (GPU) has special ray tracing support. However, most of the mature hyperspectral imaging simulation models are based on CPU implementation, but lack of GPU acceleration means.

[0041] In summary, there are technical problems such as high acquisition cost, few data sources, and low resolution in the prior art of hyperspectral remote sensing image data.

[0042] Based on this, the embodiment of the present application provides a hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation, aiming at solving the technical problems of high cost of hyperspectral remote sensing image data acquisition, few data sources and low resolution in the prior art.

[0043] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with specific embodiments and with reference to the drawings.

[0044] Figure 1 The flowchart of the hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation according to the embodiment of the present application is schematically shown.

[0045] As shown in Figure 1 The hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation of this embodiment may, for example, include operations S1-S5.

[0046] In operation S1, three-dimensional scene data is determined according to user requirements.

[0047] The three-dimensional scene data is input into a pre-constructed hyperspectral imaging simulation model, and the following operations S2-S5 are performed.

[0048] The hyperspectral imaging simulation model is configured with a global coordinate system and a three-dimensional scene object for matching the three-dimensional scene data.

[0049] In operation S2, based on the global coordinate system, a random distributed light ray is emitted from the origin to the three-dimensional scene object through a virtual imaging plane, wherein the virtual imaging plane is used to simulate the imaging surface of the remote sensor, and the three-dimensional scene object is used to simulate the optical response of the real ground object structure after the light ray is incident.

[0050] In operation S3, the intersection point of each light ray and the three-dimensional scene object is determined.

[0051] In operation S4, according to the geometric information and material information corresponding to the intersection point, combined with the atmospheric radiation parameters, the radiation data at the entrance pupil of the remote sensor is calculated.

[0052] In operation S5, the hyperspectral remote sensing picture data is generated according to the radiation data at the entrance pupil.

[0053] In this embodiment, first, the three-dimensional scene data is determined according to user requirements, which may, for example, include the geometric information and material information (such as reflectivity, scattering properties, etc.) of the ground object surface.

[0054] Then the three-dimensional scene data is input into a pre-constructed hyperspectral imaging simulation model for simulation, specifically: a light ray is projected into the scene (simulating the view angle of the remote sensor), and the interaction of the light ray with the three-dimensional scene objects (such as ground objects) in the scene is calculated, while considering the influence of the atmosphere, including atmospheric scattering, absorption, etc., the change of the light ray in the propagation process is calculated, then according to the light ray reaching the remote sensor after the influence of the atmosphere, the radiation information of the light ray is calculated and the radiation data at the entrance pupil of the remote sensor is obtained, that is, the radiance of each pixel, and finally according to the radiance calculation result, the radiance of each pixel in different wave bands is combined into a hyperspectral image.

[0055] The hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation provided in the embodiments of the present application is derived based on the ground background and target space-based imaging process, a hyperspectral imaging simulation model is constructed by using a ray tracing algorithm, through the hyperspectral imaging simulation model, light rays are emitted from the remote sensor to the three-dimensional scene in reverse for each pixel, then the propagation path and radiation information source of each light ray are tracked, the radiance information of each pixel is calculated, and finally the hyperspectral remote sensing picture data is generated. Since the hyperspectral imaging simulation model is configured with three-dimensional scene objects for simulation and simulation, the scene under various observation geometric conditions, solar conditions and atmospheric conditions can be simulated quickly according to user requirements, the qualified hyperspectral simulation data and images are generated, the requirements of various fields are met, and the economic and time costs of hyperspectral remote sensing data acquisition are significantly reduced.

[0056] According to the embodiments of the present application, the pre-constructed hyperspectral imaging simulation model can specifically include:

[0057] A local coordinate system is established with the optical center of the remote sensor as the origin, and the position and direction of the remote sensor are transformed from the local coordinate system to the global coordinate system according to the specified remote sensor observation geometric information;

[0058] Based on the global coordinate system, random distributed light rays are emitted from the origin to the three-dimensional scene objects through a virtual imaging plane; the intersection points of each light ray and the three-dimensional scene objects are determined;

[0059] According to the geometric information and material information corresponding to the intersection points, combined with the atmospheric radiation parameters, the radiation data at the entrance pupil of the remote sensor is calculated, wherein the radiation data at the entrance pupil represents the actual received radiation brightness of each pixel;

[0060] Based on the generation path of the radiation data at the entrance pupil, a hyperspectral imaging simulation model is constructed.

[0061] In the present embodiment, the construction of the hyperspectral imaging simulation model can be specifically divided into the following four steps:

[0062] (1) A coordinate system is established, as shown in Figure 2 .

[0063] Figure 2 A coordinate system diagram of a hyperspectral imaging simulation model according to an embodiment of the present application is schematically shown.

[0064] As shown in Figure 2 , first, a sensor's local coordinate system is established, the position of a pinhole in an optical system of a remote sensor (i.e. the optical center of the remote sensor) p o is located at the center of the coordinate system, a direction opposite to the observation direction of the remote sensor is defined as the e z axis of the local coordinate system, the horizontal direction and the vertical direction of the imaging plane of the remote sensor are defined as the e x axis and the e y axis of the local coordinate system.

[0065] Second, according to given remote sensor observation geometric information (for example, the remote sensor observation zenith angle θ o , the remote sensor observation azimuth angle φ o , and the remote sensor observation distance r o ), the position p o of the remote sensor, the e z axis, the e x axis and the e y axis of the local coordinate system are defined in a global coordinate system:

[0066]

[0067]

[0068]

[0069] e y =e z ×e x

[0070] (2) Projecting a light ray from the remote sensor.

[0071] Based on the position and direction of the remote sensor, a light ray is projected from the remote sensor, through each grid on a virtual imaging plane, to a three-dimensional scene object, and the light ray can be represented by the following formula:

[0072]

[0073] wherein r represents the current position of the light ray; o represents the starting point of the light ray, which is the position of the aperture stop in the imaging system of the remote sensor, i.e. p o; t represents a distance of light propagation; d represents a unit vector along a direction of light propagation, which can be calculated from a position of a pixel point on a remote sensor and a position of an imaging point of the remote sensor, and the position of the pixel of the remote sensor is p pix Therefore, d can be represented as:

[0074]

[0075] Based on a random sampling principle of a Monte Carlo method, in each pixel grid of a virtual imaging plane, a large number of randomly distributed sampling light rays are emitted from an optical center of a remote sensor to the pixel grid, so as to obtain a sufficient number of samples to ensure accuracy of Monte Carlo integration. Therefore, the position p of the pixel of the remote sensor is randomly and uniformly distributed in the pixel grid. pix

[0076] The mixed pixel effect in a remote sensing imaging process can also be truly simulated by the method.

[0077] (3) Light and scene intersection calculation.

[0078] According to the embodiment of the application, the hyperspectral imaging simulation model comprises a plurality of triangular facet elements, and determining an intersection point of each light ray and a three-dimensional scene object can specifically comprise:

[0079] determining whether the light ray intersects the triangular facet element;

[0080] in response to the light ray intersecting the triangular facet element, determining whether the intersection point is located within a boundary range of the triangular facet element;

[0081] in response to the intersection point being located within the boundary range of the triangular facet element, obtaining geometric information and material information of the intersection point.

[0082] In the embodiment, the light and scene intersection calculation is divided into the following (31) and (32) two steps:

[0083] (31) Light-plane intersection detection and intersection point calculation.

[0084] First, it is necessary to determine whether the light ray intersects a plane in which the triangular facet element is located. This step is realized by simultaneously solving a light ray expression equation and a plane equation. If there is a real solution to the simultaneous equations, the intersection point coordinates of the light ray and the plane are calculated, and it is verified whether the intersection point is located in an effective propagation direction of the light ray.

[0085] (32) Intersection point-triangular facet element containment test.

[0086] ​After confirming the intersection of the light ray and the plane, it is further necessary to judge whether the intersection point is located within the boundary range of the triangular facet. By calculating the positional relationship of the intersection point relative to the edges of the triangle, a geometric criterion is established to judge its inclusion. If the intersection point satisfies all the inclusion conditions, the three-dimensional coordinates, normal vector and other geometric information of the intersection point are returned, and the radiative characteristics are determined according to the optical properties (material information) of the triangular facet.

[0087] (4) Radiance calculation at the entrance pupil of the remote sensor.

[0088] According to the embodiments of the present application, the radiation data at the entrance pupil of the remote sensor is calculated according to the geometric information and material information corresponding to the intersection point, combined with the atmospheric radiation parameters, which can specifically include:

[0089] calculating a first ground surface reflected radiation parameter of the intersection point from the direct solar radiation;

[0090] calculating a second ground surface reflected radiation parameter of the intersection point from the scattered solar radiation;

[0091] calculating a third ground surface reflected radiation parameter of the intersection point from the background reflected radiation;

[0092] Based on the first ground surface reflected radiation parameter, the second ground surface reflected radiation parameter and the third ground surface reflected radiation parameter, the total radiation data at the entrance pupil of the remote sensor is calculated.

[0093] In the present embodiment, after obtaining the intersection point of the light ray and the scene, the ground surface reflected radiance received by the remote sensor can be calculated according to the geometric information (exit angle, incident angle, intersection normal) of the intersection point and the light ray and the material information of the intersection point, combined with the atmospheric radiation parameters (solar direct radiation irradiance reaching the ground surface, solar indirect radiation radiance reaching the ground surface, atmospheric transmittance from the ground surface to the remote sensor) during simulation.

[0094] For each light ray, the atmospheric attenuation effect during the upward process from the ground surface to the remote sensor is the same, i.e. the atmospheric transmittance is the same, so for a single light ray, only the zero-view distance reflected radiance to the ground surface needs to be calculated. The radiance at the entrance pupil of the remote sensor is obtained by statistically averaging the atmospheric transmittance of the ground surface zero-view distance reflected radiation calculated by each light fiber.

[0095] As can be seen from the rendering equation, to accurately solve the exit radiation luminance of a point P1 on the ground surface, all the incident radiation of the point needs to be calculated. The incident radiation of the intersection point has three sources: the direct radiation received by the intersection point, which can be calculated by a direct radiation sampling light ray pointing from the intersection point to the sun; the indirect radiation received by the intersection point, which comes from all incident directions ω 2 in the hemisphere space ξ iThis requires calculating the incident radiation integral over the hemispherical space.

[0096] 1. Calculation of surface reflected radiation from direct solar radiation (i.e., the first surface reflected radiation parameter).

[0097] After obtaining the geometric and material information of the intersection point P1 between the ray and the 3D scene object, it is used as a sampling point. A direct radiation sampling ray is emitted from sampling point P1 towards the sun to sample and calculate direct solar radiation. Specifically, the method involves continuously tracking the intersection of the direct radiation sampling ray with the 3D scene object. If there is no intersection with the scene, it is determined that sampling point P1 has received direct solar radiation. Figure 3 As shown.

[0098] Figure 3 The diagram illustrates the direct solar radiation received by a sampling point according to an embodiment of the present invention.

[0099] like Figure 3 As shown, at this time, the zero-line-of-sight reflected radiance L of the direct solar radiation at sampling point P1 is... d (P1,ω0) can be calculated using the following formula:

[0100]

[0101] Among them, E sd This represents the irradiance perpendicular to the direction of solar radiation reaching the Earth's surface, calculated using atmospheric radiative transfer software. ω i,d This indicates the direction of direct solar radiation, i.e., the direction of the sun. θ d This represents the angle between the direction of the sun and the direction of the normal at sampling point P1. f P1 (ω i,d , ω0 ) represents the bidirectional reflection distribution function (BRDF) of the material at sampling point P1.

[0102] If the direct radiation sampling ray intersects with a 3D scene object, then sampling point P1 is determined not to receive direct solar radiation, meaning that point is in shadow. Figure 4 As shown.

[0103] Figure 4 The diagram illustrates a sampling point that does not receive direct solar radiation according to an embodiment of the present invention.

[0104] like Figure 4 As shown, at this time, the zero-line-of-sight reflected radiance of the sampling point P1 to direct solar radiation is 0, that is:

[0105]

[0106] Second, regarding the calculation of surface reflected radiation from solar scattered radiation (i.e., the second surface reflected radiation parameter).

[0107] From the normal hemisphere space of sampling point P1 ξ 2 After emitting an indirect radiation sampling ray in a random direction, its intersection with objects in the 3D scene is tracked. If it has no intersection with the 3D scene, it is determined that the indirect radiation received by sampling point P1 comes from solar scattered radiation. Figure 5 As shown.

[0108] Figure 5 The diagram illustrates the reception of solar scattered radiation at a sampling point according to an embodiment of the present invention.

[0109] like Figure 5 As shown, at this time, the zero-line-of-sight reflected radiance L of the indirect radiation at sampling point P1 is... id (P1,ω0) can be calculated using the following formula:

[0110]

[0111] Among them, L ss (P1,ω i,ss ω represents the radiance of solar scattered radiation reaching the Earth's surface, calculated using atmospheric radiative transfer software. i,ss θ represents the direction from which solar scattered radiation originates. ss This is the angle between the direction of the solar scattered radiation source and the normal direction of the sampling point P1. (pdf) P1 (ω i,ss ) is the sampling probability density function of sampling point P1.

[0112] Third, regarding the calculation of surface reflected radiation from background reflected radiation (i.e., the third surface reflected radiation parameter).

[0113] If the indirect radiation sampling ray intersects with a 3D scene object, then the indirect radiation received by sampling point P1 is determined to originate from background reflected radiation. Figure 6 As shown.

[0114] Figure 6 The diagram illustrates the background reflected radiation received by a sampling point according to an embodiment of the present invention.

[0115] like Figure 6 As shown, at this time, the zero-line-of-sight reflected radiance L of the indirect radiation at sampling point P1 is... id (P1,ω0) can be calculated using the following formula:

[0116]

[0117] Wherein, ω i,br is the direction of the background reflected radiation source. θ br is the included angle between the direction of the background reflected radiation source and the normal direction of the sampling point P1. L br (P1, ω i ) is the background reflected radiation radiance of the surrounding background to the sampling point P1. At this time, L br (P1, ω i ) cannot be directly obtained, and the recursive idea needs to be used, and the new intersection point P2 of the indirect radiation sampling ray and the three-dimensional scene object is taken as a new sampling point for radiation sampling calculation. The sampling of the new sampling point P2 also includes direct radiation sampling and indirect radiation sampling, and if the indirect radiation sampling ray intersects the three-dimensional scene object again, the new intersection point P2 is taken as a new sampling point for radiation sampling calculation. Such recursion is performed until the indirect radiation sampling ray of the sampling point does not intersect the three-dimensional scene object, or the recursion is stopped after the number of recursions reaches a predetermined maximum threshold, and the radiation is calculated along the radiation transmission path obtained by the recursion. Generally, when the number of recursions reaches 4 times, the contribution of the background reflected radiation received by the last sampling point to the first sampling point P2 can be ignored, and at this time, the background reflected radiation received by the last sampling point is considered to be 0.

[0118] After the incident radiations of the above three kinds of source intersection points are calculated (that is, after the first ground surface reflected radiation parameter, the second ground surface reflected radiation parameter and the third ground surface reflected radiation parameter are calculated), the total radiation data at the entrance pupil of the remote sensor is calculated based on the first ground surface reflected radiation parameter, the second ground surface reflected radiation parameter and the third ground surface reflected radiation parameter.

[0119] According to the embodiments of the present application, the total radiation data at the entrance pupil of the remote sensor is calculated based on the first ground surface reflected radiation parameter, the second ground surface reflected radiation parameter and the third ground surface reflected radiation parameter, which can specifically include:

[0120] The sum of the first ground surface reflected radiation parameter and the second ground surface reflected radiation parameter is calculated, or the sum of the first ground surface reflected radiation parameter and the third ground surface reflected radiation parameter is calculated, to obtain the ground surface zero distance reflection radiance parameter of each ray;

[0121] The ground surface zero distance reflection radiance parameters of all rays are superimposed and averaged to obtain the ground surface zero distance reflection radiance parameter of each pixel;

[0122] Based on the ground surface zero distance reflection radiance parameter of each pixel, the path radiation parameter and the cross radiation are fused to obtain the total radiation data at the entrance pupil of the remote sensor.

[0123] In the present embodiment, for a single ray, the ground surface zero distance reflection radiance L d(P1, ω0) (the first ground surface reflection radiation parameter) and the ground surface zero-path-length reflection radiance L id (P1, ω0) (the second ground surface reflection radiation parameter or the third ground surface reflection radiation parameter) are added, and the ground surface zero-path-length reflection radiance L zero (P1, ω0) is:

[0124]

[0125] (P1, ω0) is: id (P1, ω0) is the second ground surface reflection radiation parameter, and the ground surface zero-path-length reflection radiance L id (P1, ω0) is the third ground surface reflection radiation parameter.

[0126] The unbiased estimation of the ground surface zero-path-length reflection radiance of each pixel is obtained by statistically averaging the ground surface zero-path-length reflection radiance calculated for each light ray. In addition, in the process of being transmitted to the entrance pupil of the remote sensor, it will be affected by atmospheric attenuation, which is represented by the uplink transmittance τ(ω0) from the ground surface point P1 to the entrance pupil P0 of the remote sensor, wherein τ(ω0) can be calculated by atmospheric radiation transmission software.

[0127] The radiation received at the entrance pupil of the remote sensor includes not only the ground surface reflection radiation, but also the path radiation L path (P1) and the intersection radiation L cross (ω0), and the radiance of the two radiations can also be calculated by atmospheric radiation transmission software. Therefore, the total radiation L ob (ω0) can be calculated by the following formula:

[0128]

[0129] (P1, ω0) represents the ground surface zero-path-length reflection radiance of the jth light ray. zero,j (P1, ω0) represents the ground surface zero-path-length reflection radiance of the jth light ray.

[0130] At this point, the hyperspectral imaging simulation model is constructed, and the hyperspectral imaging simulation model can be obtained through the paths of step (1) establishing the coordinate system, step (2) projecting the light rays from the remote sensor, step (3) light ray and scene intersection calculation, and step (4) remote sensor entrance pupil radiation calculation.

[0131] After the hyperspectral imaging simulation model is constructed, in the application of generating hyperspectral remote sensing image data by using the hyperspectral imaging simulation model, only the three-dimensional scene data determined according to the user demand is input as a model input into the hyperspectral imaging simulation model, and the generation of the hyperspectral remote sensing image data can be performed.

[0132] In this embodiment, the details of inputting the three-dimensional scene data determined according to the user demand into the hyperspectral imaging simulation model to generate the hyperspectral remote sensing image data can be referred to the construction process of the hyperspectral imaging simulation model, and the two are consistent.

[0133] For example, the details of operation S2 of emitting random distributed light rays from the origin to the three-dimensional scene object through the virtual imaging plane based on the global coordinate system are consistent with the details of step (2) of projecting the light rays from the remote sensor in the model construction process.

[0134] The details of operation S3 of determining the intersection point of each light ray and the three-dimensional scene object are consistent with the details of step (3) of calculating the intersection of the light ray and the scene in the model construction process.

[0135] And the details of operation S4 of calculating the radiation data at the entrance pupil of the remote sensor according to the geometric information and the material information corresponding to the intersection point and combining the atmospheric radiation parameters are consistent with the details of step (4) of calculating the radiation at the entrance pupil of the remote sensor in the model construction process. Therefore, details are not repeated here.

[0136] On the basis of the above embodiment, the hyperspectral imaging simulation model is implemented based on a graphics processing unit, and a rendering pipeline corresponding to a graphics application programming interface of the graphics processing unit can specifically include a path tracing renderer and a radiosity renderer.

[0137] The path tracing renderer is configured to calculate a light transmission path result by using a ray tracing model according to the input three-dimensional scene data, and output and encode the light transmission path result to a texture buffer by using a pixel shader.

[0138] The radiosity renderer is serially arranged with the path tracing renderer, and is configured to calculate the radiation data at the entrance pupil of the remote sensor according to the light transmission path result and in combination with a spectral radiation parameter.

[0139] In this embodiment, in order to accelerate the hyperspectral imaging simulation model, the rendering pipeline corresponding to the graphics application programming interface of the GPU is further improved, and specifically:

[0140] The open graphics library is a graphics application programming interface for rendering 2D and 3D vector graphics, and the core architecture thereof is a rendering pipeline of the open graphics library, which defines the entire processing flow from the input data to the output image.

[0141] The rendering pipeline of the traditional graphics application programming interface is a rasterization-based graphics rendering architecture that efficiently converts three-dimensional geometric figures into two-dimensional pixel images through a series of fixed processing stages, however, when a physical-based light tracing hyperspectral simulation needs to be implemented, the original pipeline structure must be reformed and expanded.

[0142] The present application selects to realize light tracing by deeply reforming the vertex shader and the pixel shader, which are two key color stages of the open graphics library, and also adjusts the input and output modes of the data in the pipeline. The preliminary improved graphics application programming interface light tracing RGB rendering pipeline is shown as Figure 7 .

[0143] Figure 7 The flow chart of the preliminary improved graphics application programming interface light tracing RGB rendering pipeline according to the embodiment of the present application is schematically shown.

[0144] As shown in Figure 7 , after the preliminary reform of the graphics application programming interface rendering pipeline, the core architecture of the open graphics library-based light tracing is constructed, but when it is applied to the hyperspectral remote sensing imaging simulation, there are still some challenges, for example: first, the transmission of large-scale scenes and hyperspectral data; second, the quantitative expression of radiation; third, the expansion from RGB simulation to hyperspectral simulation.

[0145] In the present embodiment, specific solutions to the three challenges will be further given.

[0146] First, regarding the transmission of large-scale scenes and hyperspectral data.

[0147] The open graphics library mainly transmits three-dimensional scene data through vertex array objects (VAO) and vertex buffer objects (VBO). This data transmission mode has obvious deficiencies when light tracing is performed, for example: VAO and VBO are used as the input source of the vertex shader, which causes the pixel shader and the pixel shader at the end of the rendering pipeline to be unable to directly read the configuration information, that is, unable to obtain the content of the scene data and the spectral data.

[0148] In view of the above two points, the embodiment selects to use a texture buffer object (TBO) to transmit the scene and hyperspectral data. The TBO can encode the three-dimensional scene geometry data, hyperspectral material data and hyperspectral radiation parameter data into a one-dimensional array in a self-defined data format in a texture map, and bind the texture map to a texture buffer. The data in the texture buffer can be directly retrieved using a texture sampler (texelFetch).

[0149] Second, regarding the quantitative expression of radiation.

[0150] The output channel of the open graphics library is a 4x2 byte array in the pixel shader, namely the fargColor array (RGBA, the A channel is generally used to represent transparency), which causes compression loss of the output radiance data, and it is difficult to meet the quantitative requirements of hyperspectral remote sensing simulation. To break through this limitation, the embodiment uses a method of encoding a single floating-point value into four channels, thereby solving the problem of insufficient quantitative accuracy of the open graphics library in hyperspectral remote sensing simulation.

[0151] Third, regarding the expansion from RGB simulation to hyperspectral simulation.

[0152] The above improvement causes the open graphics library rendering pipeline to calculate the radiance of only a single waveband in each rendering calculation. To realize simulation of hundreds of spectral wavebands, a complete ray-scene intersection operation must be performed for each waveband, which greatly affects the efficiency of large-scale hyperspectral simulation. To break through this limitation, considering that the ray paths of all wavebands in the hyperspectral simulation are completely consistent, and the amount of radiation calculation is much lower than the ray-scene intersection, the embodiment improves the open graphics library rendering pipeline as shown in Figure 8

[0153] Figure 8 A flowchart of the improved graphics application programming interface ray tracing hyperspectral rendering pipeline according to the embodiment of the application is schematically shown.

[0154] As shown in Figure 8 , the improved rendering pipeline splits the original ray tracing renderer into two serially executed renderers.

[0155] ​The first is a path tracing renderer, which only calculates the light transmission path, so only all geometry-related parameters are included in the input quantity. The path tracing renderer outputs the calculation results (the normal of each intersection of the light and the scene, the material number, the direction of the incident light and the direction of the outgoing light) through the fargColor array in the pixel shader and encodes them to the texture storage, which is used for subsequent radiation calculation. Since the information of one intersection of a light and the scene needs to store four data, one fargColor array can only output the information of one intersection. The multiple render targets (MRT) technology of the open graphics library supports simultaneously outputting the results of the shader operation to up to 8 texture maps. This technology is realized through framebuffer object (FBO). By creating up to 8 fargColor arrays and binding them to 8 FBOs respectively, up to 64 bytes of output channels can be obtained, which is enough to record all intersection information in a radiation transmission path.

[0156] The second is a radiosity renderer, which only performs radiation calculation, so only all spectrum-related parameters are included in the input quantity. The radiosity calculation renderer performs simulation calculation of one spectral band per frame, takes the radiation path information calculated by the path tracing renderer as input, combines the spectral radiation information input by itself, calculates the spectral radiance at the entrance pupil of the remote sensor, and finally outputs and stores the hyperspectral remote sensing simulation image.

[0157] The hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation provided by the embodiment of the application uses GPU to realize acceleration of the hyperspectral imaging simulation model, that is, by improving the rendering pipeline corresponding to the graphics application programming interface of the graphics processing unit, introducing a path tracing renderer and a radiosity renderer, realizing hyperspectral rendering, and solving the data input and output problem, the quantification problem and the multi-spectral rendering problem of the original rendering pipeline. Since the improved rendering pipeline splits the original ray tracing renderer into two serially executed renderers, the amount of radiation calculation is greatly reduced, and the efficiency of large-scale hyperspectral simulation is improved.

[0158] On the basis of the above-mentioned embodiment, in order to further reduce resource consumption and running time, the path tracing renderer is further accelerated.

[0159] Even after the rendering pipeline of the graphics application programming interface is improved, the ray tracing algorithm still needs a large amount of computing resources, and the long running time and resource consumption hinder the practical application of the hyperspectral imaging simulation model. To solve this problem, the application proposes an acceleration method under the premise of maintaining the required accuracy standard, which implements the following acceleration optimization operations while meeting the accuracy requirements.

[0160] According to an embodiment of the present application, the path tracing renderer is configured to accelerate the ray tracing model by means of Monte Carlo random sampling, wherein the path tracing renderer comprises a hierarchical bounding volume model, an importance sampling module, a low-discrepancy sequence sampling module and a simulation accuracy and simulation efficiency balancing module.

[0161] The hierarchical bounding volume model is used to accelerate the detection of the intersection feature of the light ray and the three-dimensional scene object, wherein the hierarchical bounding volume model is configured to search the intersection feature of the light ray and the bounding box of each node in turn from the root node according to a depth-first search strategy.

[0162] The hierarchical bounding volume structure (BVH) is a high-efficiency data structure based on object space division. By using the principle of spatial aggregation characteristics of geometric element distribution in a three-dimensional scene, a progressive screening mechanism from the whole to the local is formed by constructing a hierarchical bounding volume structure, and unnecessary geometric primitive intersection tests are greatly reduced.

[0163] The typical BVH construction process adopts a recursive algorithm, that is, first starting from the root node containing all primitives, and then repeatedly performing the following operations: calculating the bounding box of the current node, selecting the best split axis, determining the split position, and generating left and right child nodes. This recursive process continues until the preset termination condition is met (such as the number of primitives contained by the node reaching the lower limit or reaching the maximum recursion depth).

[0164] In the present embodiment, when the ray tracing is applied, the traversal of the BVH adopts a depth-first search strategy, that is, the algorithm starts from the root node and detects the intersection of the light ray and the bounding box of each node in turn. For the node intersecting with the light ray, all the triangular face elements contained therein are further checked; while the non-intersecting node and its child nodes are directly skipped. When the light ray passes through multiple child nodes at the same time, the algorithm uses data structures such as priority queues to preferentially process the node closest to the starting point of the light ray. This "nearest first" processing strategy can quickly locate the nearest intersection point, significantly improving the algorithm efficiency.

[0165] The importance sampling module is used to preferentially select the direction or target with the greatest impact on the result for sampling in the Monte Carlo random sampling process.

[0166] In the present embodiment, the Monte Carlo method is an algorithm for solving problems by estimating the probability of a random event through random sampling. If the Monte Carlo method is used to calculate the integral of a function, let X1, X2…X n are independent samples and subject to the same distribution, and the probability density function is denoted as pdf(x), then the integral of the function can be expressed as:

[0167]

[0168] The variance of the estimation method decreases with the increase of the sample number, that is, more accurate. The importance sampling method is a method for increasing accuracy and reducing variance while keeping the sample number constant. The principle is as follows: let the standard deviation of the Monte Carlo estimator be σ, and it can be known that:

[0169]

[0170] If the variance of the Monte Carlo integral is to be minimized, then is needed, which is equivalent to , that is, when Therefore, importance sampling is to sample more in the region where the original function contributes more to the integral, so as to obtain more accurate data. At this time, the sampling is not uniformly distributed, so each sampling value needs to be weighted, and this weight is called importance weight.

[0171] The original Monte Carlo ray tracing uses a half-sphere uniform distribution sampling probability density function, that is:

[0172]

[0173] This sampling distribution does not satisfy . Since there is a fixed cosine factor in the Monte Carlo integral formula of the hemisphere radiation, it is used as the importance weight, and the sampling distribution is adjusted to the cosine distribution, that is:

[0174]

[0175] At this time, the sampling distribution satisfies , and the precision is the highest under the same sample number.

[0176] The low-discrepancy sequence sampling module is used to generate a uniformly distributed point set in the sample space through mathematical construction, to improve the convergence speed of the Monte Carlo integral.

[0177] In the Monte Carlo integral radiation sampling of ray tracing, an important factor affecting the convergence speed of the integral is the quality of the random number sequence. However, the random number sequence may appear uneven distribution in some cases, resulting in slow convergence speed. As a scheme for generating random numbers, low-discrepancy sequences can significantly improve the convergence speed of the Monte Carlo integral by mathematically constructing a more uniformly distributed point set in the sample space.

[0178] In the actual ray tracing simulation calculation in the embodiment, the low-discrepancy sequence is usually used in combination with importance sampling. The importance sampling adjusts the sampling density function to make it similar to the shape of the integrand function, and the low-discrepancy sequence ensures that the sampling points are still well-distributed in the adjusted space.

[0179] The simulation accuracy and simulation efficiency balancing module is configured to determine an optimal number of rays according to the simulation accuracy and the simulation efficiency.

[0180] Although the increase in the number of rays determines the calculation accuracy, the marginal benefit will gradually decrease as the number of rays increases. The complexity T of ray tracing is positively correlated with the number n of rays, as shown in the following formula:

[0181]

[0182] Therefore, the embodiment explores the relationship between the number of rays and the calculation accuracy from the perspective of balancing the calculation efficiency and the calculation accuracy, and attempts to find an optimal number of rays. The simulation results under different numbers of rays are statistically analyzed based on the simulation results of 10,000 rays as the benchmark image. The comparison parameters include the mean and variance of the relative difference of the pixel-by-pixel radiant luminance. Finally, the model selects 2,000 rays as the best balance point between efficiency and calculation accuracy. Compared with 10,000 rays, this optimization achieves a 70.4% improvement in calculation efficiency with only an acceptable 0.16% loss in accuracy.

[0183] The hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation provided by the embodiment of the application accelerates the intersection of rays and scenes using hierarchical bounding boxes, accelerates the convergence of Monte Carlo integration using importance sampling and low-discrepancy sequences, and selects an optimal number of rays for simulation experiments through exploration of the relationship between the number of rays and the simulation accuracy, thereby achieving a substantial improvement in simulation efficiency at a very small accuracy cost.

[0184] On the basis of the above embodiment, in order to better put the hyperspectral imaging simulation model into actual hyperspectral imaging simulation applications, the embodiment is based on the improved open graphics library ray tracing rendering pipeline, and a scene editing module based on the BS (Browser / Serve) architecture is used as the front end to further develop a hyperspectral remote sensing imaging simulation software. The simulation process and main function modules of the software are as shown in Figure 9 .

[0185] Figure 9 The hyperspectral remote sensing imaging simulation software process and function module block diagram according to the embodiment of the application are schematically shown.

[0186] As Figure 9As shown, the hyperspectral remote sensing imaging simulation software process of the embodiment of the application may, for example, include:

[0187] (1) Open the software, start the simulation system, and enter the main interface of the software.

[0188] (2) New scene: the user starts to build a new simulation task scene, which usually includes specifying ground objects, terrain, buildings, sensors, etc.

[0189] (3) Scene editing: import or edit three-dimensional model and material information: including background, target objects, etc.; finally output a complete scene combination package (geometry + material).

[0190] (4) New calculation task: a new simulation calculation task is created for the scene, and the task may include: simulation type (hyperspectral / multispectral), sensor parameters, solar elevation angle, waveband setting, etc.

[0191] (5) Parameter configuration: the user inputs or loads an existing parameter configuration file, and the configuration items include: light tracing settings (precision, sampling rate, number of light rays, etc.); material reflectivity, BRDF model; waveband range, spectral resolution; environmental conditions (atmospheric model, solar angle); and loading of material reflectivity files (for hyperspectral reflection calculation).

[0192] (6) Simulation calculation: start the light tracing engine for rendering and hyperspectral inversion simulation; the acceleration module is applied: level bounding box acceleration intersection; importance sampling optimizes light direction; low difference sequence improves convergence efficiency; precision control reduces the number of light rays; and the output is a hyperspectral simulation data cube or image.

[0193] (7) Output simulation results, including hyperspectral images, waveband images, spectral curves, etc.; which can be used for subsequent imaging analysis, sensor design verification or target recognition simulation, etc.

[0194] The embodiment of the application also provides a hyperspectral remote sensing image generation device based on three-dimensional scene radiation simulation, which comprises a first determination module, an input module, an emission module, a second determination module, a calculation module and a generation module.

[0195] The first determination module is configured to determine three-dimensional scene data according to user requirements.

[0196] The input module is configured to input the three-dimensional scene data into a pre-constructed hyperspectral imaging simulation model, wherein the hyperspectral imaging simulation model is configured with a global coordinate system and a three-dimensional scene object for matching the three-dimensional scene data, and the following modules are executed:

[0197] The emission module is used to emit randomly distributed light rays from the origin to the three-dimensional scene object based on the global coordinate system and through the virtual imaging plane. The virtual imaging plane is used to simulate the imaging surface of the remote sensor, and the three-dimensional scene object is used to simulate the optical response of the real ground structure after the light rays are incident.

[0198] The second determining module is used to determine the intersection point of each ray with the three-dimensional scene object.

[0199] The calculation module is used to calculate the radiation data at the entrance pupil of the remote sensor based on the geometric and material information corresponding to the intersection point, combined with atmospheric radiation parameters.

[0200] The generation module is used to generate hyperspectral remote sensing image data based on the radiation data at the entrance pupil.

[0201] Figure 10 The diagram illustrates the structure of an electronic device suitable for implementing a hyperspectral remote sensing image generation method based on three-dimensional scene radiation simulation according to an embodiment of the present invention. Figure 10 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0202] like Figure 10 As shown, an electronic device 1000 according to an embodiment of the present invention includes a processor 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage portion 1008 into a random access memory (RAM) 1003. The processor 1001 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 1001 may also include onboard memory for caching purposes. The processor 1001 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0203] The storage section 1008 stores various programs and data required for the operation of the electronic device 1000. The processor 1001, ROM 1002, and storage section 1008 are interconnected via bus 1004. The processor 1001 executes various operations of the method flow according to embodiments of the present invention by executing programs in ROM 1002 and / or storage section 1008. It should be noted that the programs may also be stored in one or more memories other than ROM 1002 and storage section 1008. The processor 1001 may also execute various operations of the method flow according to embodiments of the present invention by executing programs stored in said one or more memories.

[0204] According to an embodiment of the present application, the electronic device 1000 can further include an input / output (I / O) interface 1005 that is also connected to the bus 1004. The electronic device 1000 can further include one or more of the following components connected to the input / output (I / O) interface 1005: an input part 1006 including a keyboard, a mouse, etc.; an output part 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage part 1008 including a hard disk, etc.; and a communication part 1009 including a network interface card such as a LAN card, a modem, etc. The communication part 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the input / output (I / O) interface 1005 as necessary. A removable medium 1011 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is mounted on the drive 1010 as necessary, so that a computer program read out therefrom is installed in the storage part 1008 as necessary.

[0205] According to an embodiment of the present application, the method flow according to the embodiment of the present application can be implemented as a computer software program. For example, the embodiment of the present application includes a computer program product including a computer program carried on a computer-readable storage medium, the computer program containing program codes for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network by the communication part 1009, and / or installed from the removable medium 1011. When the computer program is executed by the processor 1001, the above-described functions defined in the system implementing the embodiment of the present application are executed. According to an embodiment of the present application, the system, device, apparatus, module, unit, etc. described above can be implemented by computer program modules.

[0206] The present application also provides a computer-readable storage medium, which can be included in the device / apparatus / system described in the above embodiments; or can exist separately without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiment of the present application.

[0207] According to an embodiment of the present application, the computer readable storage medium can be a non-transitory computer readable storage medium. For example, it can include, but is not limited to, portable computer disks, hard disks, random access memories (RAM), read only memories (ROM), erasable programmable read only memories (EPROM or flash memory), portable compact disk read only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. In this application, the computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0208] For example, according to an embodiment of the present application, the computer readable storage medium can include the ROM 1002 and / or the storage portion 1008 described above, and / or one or more memories other than the ROM 1002 and the storage portion 1008.

[0209] The flow diagrams and the block diagrams in the drawings are illustrations of possible architectures, functions, and operations according to various embodiments of the present application. In this regard, each block in the flow diagrams or block diagrams can represent a module, a segment, or a portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that in some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flow diagrams, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or operations, or combinations of special purpose hardware and computer instructions.

[0210] Embodiments of the present application have been described above. However, these embodiments are merely meant to be illustrative and not limiting of the scope of the present application. Although each embodiment has been described separately above, this does not mean that measures in the various embodiments cannot be used advantageously in combination. Numerous alternatives and modifications of embodiments of the present application can be devised by those skilled in the art without departing from the inventive concept as set out in the following claims.

Claims

1. A method for generating hyperspectral remote sensing images based on three-dimensional scene radiation simulation, characterized in that, The method includes: Determine the 3D scene data based on user requirements; The three-dimensional scene data is input into a pre-constructed hyperspectral imaging simulation model, wherein the hyperspectral imaging simulation model is configured with a global coordinate system and a three-dimensional scene object for matching the three-dimensional scene data, and performs the following operations: Based on the global coordinate system, starting from the origin, randomly distributed light rays are emitted towards the three-dimensional scene object through a virtual imaging plane. The virtual imaging plane is used to simulate the imaging surface of a remote sensor, and the three-dimensional scene object is used to simulate the optical response of real ground structures after light is incident. Determine the intersection point of each ray with the 3D scene object; Based on the geometric and material information corresponding to the intersection point, and combined with atmospheric radiation parameters, the radiation data at the entrance pupil of the remote sensor is calculated, including: Calculate the first surface reflected radiation parameters of the intersection point from direct solar radiation; Calculate the second surface reflected radiation parameters of the intersection point from solar scattered radiation; Calculate the third surface reflected radiation parameter of the intersection point from the background reflected radiation; Based on the first surface reflected radiation parameter, the second surface reflected radiation parameter, and the third surface reflected radiation parameter, the total radiation data at the entrance pupil of the remote sensor is calculated. Based on the radiation data at the entrance pupil, hyperspectral remote sensing image data is generated; The hyperspectral imaging simulation model is implemented based on a graphics processing unit (GPU), wherein the rendering pipeline corresponding to the GPU's graphics application programming interface includes: The path tracing renderer is used to calculate the ray transmission path results based on the input 3D scene data using a ray tracing model, and output the ray transmission path results through a pixel shader and encode them into a texture buffer for storage. A radiometric renderer, configured in series with the path tracing renderer, is used to calculate the radiation data at the entrance pupil of the remote sensor based on the light transmission path results and spectral radiation parameters.

2. The method according to claim 1, characterized in that, The hyperspectral imaging simulation model includes multiple triangular primitives, wherein determining the intersection point of each ray with the three-dimensional scene object includes: Determine whether the light ray intersects with the triangular element surface; In response to the intersection of the light ray with the triangular element face, determine whether the intersection point is located within the boundary range of the triangular element face; In response to the intersection point being located within the boundary range of the triangular element face, the geometric and material information of the intersection point is obtained.

3. The method according to claim 1, characterized in that, The calculation of the total radiation data at the entrance pupil of the remote sensor based on the first surface reflected radiation parameter, the second surface reflected radiation parameter, and the third surface reflected radiation parameter includes: Calculate the sum of the first surface reflected radiation parameter and the second surface reflected radiation parameter, or calculate the sum of the first surface reflected radiation parameter and the third surface reflected radiation parameter, to obtain the surface zero-line-of-sight reflected radiance parameter for each ray. The surface zero-line-of-sight reflectance radiance parameters of all light rays are superimposed and the average value is calculated to obtain the surface zero-line-of-sight reflectance radiance parameters of each pixel. Based on the zero-line-of-sight reflection radiance parameter of each pixel, the path radiation parameter and cross radiation are fused to obtain the total radiation data at the entrance pupil of the remote sensor.

4. The method according to claim 1, characterized in that, The pre-construction of the hyperspectral imaging simulation model includes: A local coordinate system is established with the optical center of the remote sensor as the origin, and the position and orientation of the remote sensor are transformed from the local coordinate system to the global coordinate system based on the specified observation geometry information of the remote sensor. Based on the global coordinate system, starting from the origin, randomly distributed rays are emitted from the virtual imaging plane toward the three-dimensional scene object; Determine the intersection point of each ray with the 3D scene object; Based on the geometric and material information corresponding to the intersection point, and combined with atmospheric radiation parameters, the radiation data at the entrance pupil of the remote sensor is calculated, wherein the radiation data at the entrance pupil represents the actual radiation brightness received by each pixel; The hyperspectral imaging simulation model is constructed based on the generation path of the radiation data at the entrance pupil.

5. The method according to claim 1, characterized in that, The path tracing renderer is configured to accelerate the ray tracing model using Monte Carlo random sampling, wherein the path tracing renderer includes: A hierarchical bounding box model is used to accelerate the detection of intersection features between light rays and the three-dimensional scene objects. The hierarchical bounding box model is configured to search for intersection features between light rays and the bounding boxes of each node sequentially, starting from the root node, according to a depth-first search strategy. The importance sampling module is used to prioritize sampling the direction or target that has the greatest impact on the result during the Monte Carlo random sampling process; The low-difference sequence sampling module is used to generate a uniformly distributed set of points in the sample space through mathematical construction, thereby improving the convergence speed of Monte Carlo integrals. The simulation accuracy and efficiency balancing module is used to determine the optimal number of rays based on simulation accuracy and efficiency.

6. A hyperspectral remote sensing image generation device based on three-dimensional scene radiation simulation, characterized in that, The device includes: The first determination module is used to determine the 3D scene data according to user requirements; An input module is used to input the three-dimensional scene data into a pre-constructed hyperspectral imaging simulation model, wherein the hyperspectral imaging simulation model is configured with a global coordinate system and a three-dimensional scene object for matching the three-dimensional scene data, and executes the following modules: The emission module is used to emit randomly distributed light rays from the origin to the three-dimensional scene object based on the global coordinate system, through a virtual imaging plane. The virtual imaging plane is used to simulate the imaging surface of a remote sensor, and the three-dimensional scene object is used to simulate the optical response of real ground structures after light is incident. The second determining module is used to determine the intersection point of each ray with the three-dimensional scene object; The calculation module is used to calculate the radiation data at the entrance pupil of the remote sensor based on the geometric and material information corresponding to the intersection point, combined with atmospheric radiation parameters, including: Calculate the first surface reflected radiation parameters of the intersection point from direct solar radiation; Calculate the second surface reflected radiation parameters of the intersection point from solar scattered radiation; Calculate the third surface reflected radiation parameter of the intersection point from the background reflected radiation; Based on the first surface reflected radiation parameter, the second surface reflected radiation parameter, and the third surface reflected radiation parameter, the total radiation data at the entrance pupil of the remote sensor is calculated. The generation module is used to generate hyperspectral remote sensing image data based on the radiation data at the entrance pupil; The hyperspectral imaging simulation model is implemented based on a graphics processing unit (GPU), wherein the rendering pipeline corresponding to the GPU's graphics application programming interface includes: The path tracing renderer is used to calculate the ray transmission path results based on the input 3D scene data using a ray tracing model, and output the ray transmission path results through a pixel shader and encode them into a texture buffer for storage. A radiometric renderer, configured in series with the path tracing renderer, is used to calculate the radiation data at the entrance pupil of the remote sensor based on the light transmission path results and spectral radiation parameters.

7. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors perform the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having executable instructions stored thereon, characterized in that, When this instruction is executed by the processor, it causes the processor to perform the method according to any one of claims 1 to 5.

Images