Three-dimensional radiance calculation method for high-resolution remote sensing scene and application

Abstract

The invention relates to the technical field of space remote sensing, and particularly provides a three-dimensional radiance calculation method for a high-resolution remote sensing scene and application thereof.The method comprises the steps that firstly, a surface element grid of a three-dimensional space is created, spontaneous radiation of all triangular surface elements under sunlight irradiation is calculated, then a traditional RGM model is improved, and under an OpenACC parallel framework, the three-dimensional radiance of the high-resolution remote sensing scene is calculated; visual factors are parallelly calculated through multiple threads of a GPU, then the radiation flux density of surface elements is iteratively calculated by adopting a successive super-relaxation method, and after the radiation flux density of each surface element is obtained, the bidirectional reflectivity BRF at different observation angles is calculated, so that efficient simulation of the bidirectional reflectivity and radiation transmission of a complex scene is realized. According to the method, the calculation speed of an original RGM model is remarkably improved, the calculation speed is improved by nearly 70 times through testing, and when a large-scale scene is simulated, ten-million-level small patches can be processed without sacrificing precision.

Description

Technical Field

[0001] The invention belongs to the field of space remote sensing technology, and in particular relates to a three-dimensional radiometry calculation method for a high-resolution remote sensing scene and its application. Background Art

[0002] With the rapid development of remote sensing technology, the importance of three-dimensional radiative transfer (RT) models in simulating bidirectional reflectance factors (BRFs) has become increasingly prominent. Compared with one-dimensional RT models, three-dimensional RT models have played an important role in applications such as forest structure and biophysical parameter extraction because they rely less on assumptions and can provide higher accuracy. They are also widely used in the study of vegetation, forests, and urban scenes. RT models can accurately calculate multiple scattering and maintain energy conservation. Commonly used models include discrete coordinate method (DOM), Monte Carlo radiative transfer method (MCRT), and radiometric model (RGM).

[0003] The RGM model has been widely used in the field of vegetation remote sensing and is suitable for tasks such as plant index design, leaf size and shape extraction, and scale effect research. However, the computational efficiency of the RGM model is low, especially when simulating large heterogeneous canopy scenes at the pixel scale. The calculation of its field of view factor often faces the limitation of inefficiency or even inability to complete. This problem hinders its extended application in high-resolution and larger scenes. For example, when the scene contains a complex structure with more than 100,000 faces, the calculation of the field of view factor may make the simulation process impossible.

[0004] Although methods such as the RAPID model improve computational efficiency by simplifying the scene, excessive simplification can lead to reduced accuracy of simulation results. In heterogeneous canopy scenes, preserving the fine structure of the scene is crucial to the accuracy of BRF simulation. Therefore, there is an urgent need for a solution that can efficiently handle the field of view factor calculation while preserving the details of the canopy scene as much as possible to meet the simulation needs of high-resolution remote sensing scenes at the pixel scale. Summary of the invention

[0005] In view of this, the present invention aims to provide a three-dimensional radiometry calculation method and application for high-resolution remote sensing scenes, improve the traditional radiometry model (RGM) based on the OpenACC parallel framework, optimize and rewrite the field of view factor calculation part of the RGM model through OpenACC technology, and optimize the polygon filling algorithm to fill the area of ​​the surface element, improve the calculation equation of the surface element radiation flux density, break through the efficiency limitation of the field of view factor calculation, and thus realize high-precision BRF simulation at the pixel scale.

[0006] To achieve the above object, the technical solution created by the present invention is implemented as follows: The present invention provides a method for calculating three-dimensional radiometry of a high-resolution remote sensing scene, comprising: generating a triangular facet mesh by defining vertices in a three-dimensional space to create a three-dimensional scene; Calculate the flux density of spontaneous radiation of the surface element under sunlight based on the optical parameters of the surface element; Calculating the visual factor through the OpenACC parallel framework, including: transferring the face data used to calculate the visual factor from the CPU to the GPU, allocating the calculation tasks of the visual factor loop iteration between different facets to multiple threads of the GPU, and accelerating the visual factor calculation; The radiation flux density of the surface element is iteratively calculated using the successive super-relaxation method. The iterative calculation formula is: ; in, represents the relaxation factor, represents the k+1th iteration radiation flux density of the i-th surface element, represents the k-th iteration radiation flux density of the i-th surface element, represents the k-th iteration radiation flux density of the j-th surface element, represents the reflectivity or transmittance of the jth face element, N represents the total number of face elements, represents the visibility factor from the jth bin to the ith bin, represents the flux density of spontaneous radiation of the i-th surface element under sunlight; The bidirectional reflectivity BRF at different observation angles is calculated based on the radiation flux density of each surface element. The calculation formula is: ; in, represents the observation angle, Indicated in Bidirectional reflectivity at angles, represents the normal direction of the ith face element, From the observation point of view The area of ​​the i-th face element seen, Represents the radiation flux density of the i-th surface element.

[0007] Preferably, the calculation formula for the flux density of spontaneous radiation of the surface element under sunlight is: ; in, represents the area of ​​the ith facet, Represents any point on the i-th face The flux density of spontaneous emission.

[0008] Preferably, the calculation formula of the visual factor is: ; in, Express differential, Express differential, express and Connection between The angle of the surface normal, express and Connection between The angle of the surface normal, express and The Euclidean distance between .

[0009] Preferably, under the OpenACC parallel framework, at least one strategy among parallel loop instructions, optimized data layout, minimized data transfer and dynamic memory allocation is adopted to accelerate the field of view factor calculation.

[0010] Preferably, the polygon filling algorithm is used to fill the area of ​​the face element, and the Cartesian coordinates of any point in the three-dimensional space are converted into barycentric coordinates, and the barycentric coordinates are used to determine whether the point is located in a certain face element.

[0011] Preferably, in the polygon filling algorithm, the coordinates of the three vertices A, B and C of the face element are used to represent any point P as: ; in, and represents the centroid coordinates of point P; when and If both are non-negative numbers, then point P is inside the bin; otherwise, point P is outside the bin.

[0012] Preferably, the barycentric coordinates and The calculation formula is: ; in, , and Represents vectors , and The component in the x direction, , and Represents vectors , and Component in the y direction.

[0013] Preferably, during the visualization factor calculation process, the storage structure in the GPU is a hash-based data structure.

[0014] Preferably, the range of the relaxation factor is 0 to 2, and in the process of solving the radiation flux density, the optimal relaxation factor is searched within the range of the relaxation factor with a step size of 0.2.

[0015] On the other hand, the invention provides an application of a three-dimensional radiometry calculation method for a high-resolution remote sensing scene, uses the three-dimensional radiometry calculation method for a high-resolution remote sensing scene to construct a GPU-based three-dimensional radiometry graphics model, and uses the three-dimensional radiometry graphics model for heterogeneous canopy simulation.

[0016] Compared with the prior art, the invention can achieve the following beneficial effects: The present invention improves the traditional radiosity model (RGM) based on the OpenACC parallel framework, and proposes a three-dimensional radiosity calculation method with higher computational efficiency, larger computational scale, and guaranteed modeling accuracy. The present invention transfers data calculation from the CPU to the GPU, and designs an improved radiosity model RGM using the OpenACC parallel technology in the GPU. PS Through parallel loop instructions, optimized data layout, minimized data transfer and dynamic memory allocation strategies, it has been tested that the calculation speed of the visual factor has been increased by about 70 times. It can process tens of millions of small patches without sacrificing accuracy, greatly improving the simulation capability of large-scale scenes and solving the problem of low calculation efficiency of the original RGM model and algorithm.

[0017] The present invention has good accuracy and reliability in simulating the bidirectional reflectivity and radiation transmission of complex scenes. For example, in the simulation test of complex canopy, the improved radiosity model RGM constructed by the method of the present invention is PS The correlation (R²) with the original RGM model exceeded 0.94, the root mean square error (RMSE) was lower than 0.0038, and cross-validation with the DART model showed high consistency (R² as high as 0.98), with RMSEs of 0.0031 and 0.0340 in the red and near-infrared bands, respectively, showing extremely high accuracy and reliability.

[0018] The parallel iterative operation of the present invention not only significantly improves the calculation speed, but also expands the computable scale, increasing the complexity of the simulated scene from tens of thousands of faces of the original RGM model to 14 million faces, which can meet the simulation needs of larger-scale and high-resolution remote sensing scenes, and provides the possibility for studying larger-scale and more complex ecosystems. It is currently applicable to fields such as vegetation remote sensing, forest structure and biophysical parameter extraction, and urban scene research, and can achieve more efficient and accurate simulation.

[0019] The present invention optimizes the traditional methods of face filling and projection area calculation, designs a triangle filling algorithm (TFBCs) based on barycentric coordinates, and replaces the traditional line scanning algorithm with a triangle filling algorithm based on barycentric coordinates, thereby improving the efficiency of filling and projection area calculation, being more suitable for GPU acceleration, and further improving the computing performance of the model. By optimizing the polygon filling algorithm and parallel strategy, it is ensured that the model exhibits high precision in the BRF simulation of both the main plane and the intersection plane, and has a good correlation with the DART model, achieving a balance between precision and efficiency.

[0020] The present invention transforms the data storage structure in the simulation calculation process from a traditional array to a hash-based data structure, improves the data query efficiency from O(n) to O(1), and speeds up the access and processing of data in the visual factor calculation process. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The drawings constituting part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments and descriptions of the present invention are used to explain the present invention and do not constitute an improper limitation on the present invention. In the drawings: Figure 1 is a flow chart of a method for calculating three-dimensional radiosity of a high-resolution remote sensing scene provided in an embodiment of the present invention; Figure 2 is an operation strategy diagram for parallel processing of visual factors under the OpenACC parallel framework provided in an embodiment of the present invention; Figure 3 The DART model provided by the embodiment of the present invention and the RGM constructed by the method of the present invention are PS Comparison results of simulated reflectivity; Figure 4 is an RGM provided according to an embodiment of the present invention PS Comparison of simulation results of DART and 4SAIL models with experimental results of real-scene airborne measurements. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the invention clearer, the invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the invention and do not constitute a limitation to the invention. Similar components in different embodiments use associated similar component numbers. In the following embodiments, many detailed descriptions are to enable the invention to be better understood. However, those skilled in the art can easily recognize that some of the features can be omitted in different situations, or can be replaced by other components, materials, and methods. In some cases, some operations related to the invention are not shown or described in the specification, in order to avoid the core part of the invention being overwhelmed by too much description, and for those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations according to the description in the specification and the general technical knowledge in the art.

[0023] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to form various implementation methods. At the same time, the steps or actions in the method description can also be interchanged or adjusted in a manner that is obvious to those skilled in the art. Therefore, the various sequences in the specification and the drawings are only for the purpose of clearly describing a certain embodiment and are not meant to be a necessary sequence, unless otherwise specified that a certain sequence must be followed.

[0024] In the description of the invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing the invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the invention. In addition, the terms "first", "second", etc. are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first", "second", etc. may explicitly or implicitly include one or more of the features. In the description of the invention, unless otherwise specified, the meaning of "multiple" is two or more.

[0025] In the description of the invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installation", "connection" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the invention can be understood according to specific circumstances.

[0026] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0027] See also Figure 1 In one embodiment of the present invention, a method for calculating the three-dimensional radiosity of a high-resolution remote sensing scene is provided. The method is used to optimize and improve the traditional radiosity model (RGM). The three-dimensional radiosity improvement model (RGM) is constructed by using the OpenACC parallel technology in a GPU environment. PS ), which breaks through the efficiency limitation of the traditional RGM model in calculating the field of view factor, and can realize the efficient simulation of the bidirectional reflectivity and radiation transmission of complex canopies. The three-dimensional radiosity calculation method is as follows: S1: First, create a three-dimensional scene. Specifically, in three-dimensional space, a triangular mesh is generated by defining a series of vertices, and complex geometric shapes are decomposed into multiple triangles. Each triangle is a face element, and each triangular face element is determined by three vertices.

[0028] These triangular face elements can be combined into complex geometric shapes, such as vegetation leaves, buildings and terrain. Correspondingly, the outer vertices of the figures composed of these face elements can represent the corner points of complex geometric shapes such as vegetation leaves, buildings and terrain.

[0029] Spectral properties such as reflectivity and transmittance are defined for the surface of each triangular facet to achieve the definition of scene surface properties. These spectral properties will be used for subsequent radiosity calculations.

[0030] S2: After the three-dimensional scene is created, it is necessary to determine the emittance of all facets in the three-dimensional scene, that is, the flux density of spontaneous radiation generated by each facet under sunlight. Specifically, assuming that the total number of facets in the three-dimensional scene is N, the radiation flux density of each facet will be affected by solar radiation on the one hand, and by the radiation of other facets on the other hand. Specifically, the radiance of any discrete triangular facet i can be expressed as: ; in, represents the area of ​​the ith facet, represents the area of ​​the jth facet, Express Differential, that is, the differential area of ​​surface element i, represents the radiation flux density of the i-th surface element, represents the flux density of spontaneous radiation of the i-th surface element under sunlight; represents the reflectivity or transmittance of the ith surface element, that is, the optical parameter of the surface element, Represents the visual factor of the differential area of ​​the jth bin to the differential area of ​​the ith bin. It should be noted that the value of i can be 2N, because it is necessary to consider the front and back sides of N bins.

[0031] Therefore, in this step, the flux density of spontaneous radiation generated by the surface element under sunlight can be calculated first. The flux density of spontaneous radiation generated by any i-th surface element under sunlight can be solved by the following formula: ; in, Represents any point on the i-th face The flux density of spontaneous emission.

[0032] The visible factor in the radiosity calculation formula is a basic parameter in radiation transfer analysis, which quantifies the The radiation flux density emitted reaches the infinitesimal surface The ratio of , so the calculation formula of the visual factor is: ; in, express and Connection between The angle of the surface normal, express and Connection between The angle of the surface normal, express and The Euclidean distance between .

[0033] Through the above, the visible factor of the face element after differentiation can be determined, and then the visible factor between the face elements can be obtained by integration. However, various models and methods currently mainly use the ray projection method, use the line scanning algorithm to fill the area and the subsequent projection area method for calculation. The calculation efficiency of the visible factor between the huge number of face elements in the three-dimensional scene is extremely low. The calculation of the visible factor often faces the problem of low efficiency or even incompletion. The low efficiency problem makes it impossible to realize the three-dimensional radiosity calculation of high-resolution or larger scenes, which limits the existing models and methods to only process scenes with no more than 100,000 face elements. The number of face elements in high-resolution and large scenes usually needs to be greater than 100,000 face elements. If it is compressed to reduce the number of face elements, it will affect the simulation accuracy, thus greatly limiting the application of the RGM model.

[0034] S3: Since the visual factor calculation occupies about 85% of the total calculation time of remote sensing simulation, the present invention focuses on optimizing the iterative calculation process of the visual factor, and uses OpenACC parallel acceleration technology to speed up the running time without affecting the calculation accuracy. Specifically, the OpenACC parallel framework is used to calculate the visual factor. The OpenACC parallel framework can transfer the calculation data and calculation process from the CPU to the GPU, and use the powerful computing power of the GPU to distribute the calculation tasks of the cyclic iteration of the visual factor between different face elements to multiple threads of the GPU, thereby accelerating the calculation of the visual factor. The OpenACC framework aims to promote code development and simplify the porting of existing codes on heterogeneous computing systems to obtain better parallel performance. In addition, the present invention also adopts an incremental method and a combination of multiple strategies to improve RGM. PS The correctness and processing speed of the model under the OpenACC framework.

[0035] For specific acceleration strategies, please refer to Figure 2 ,include: OpenACC's parallel loop directives are used to distribute loop iterations to multiple threads, thereby speeding up loop execution. The number of threads launched per GPU core function and their clustering can be fine-grained optimized using vector, worker, and gang directives.

[0036] Compared with the array of structures (AOS) scheme in the CPU working mode, the GPU working mode is more suitable for the array structure (SoA) scheme to meet the requirements of merged memory access in the RGMPS implementation. The implementation of the distribution function index formula requires the array Convert to To adapt to the single instruction multiple threads (SIMTs) execution model. Therefore, the data layout of the facet metadata and the computing data needs to be optimized. The data layout can be optimized through data compression, data alignment, and data packing technologies to arrange the data in a way that maximizes performance and minimizes memory usage.

[0037] For data transfer between the CPU and GPU, data transfer optimization is also minimized, which is achieved by optimizing the code to reduce the amount of data that needs to be processed or using more efficient data transfer methods. Specifically, the copyin clause copies the array in the CPU cache or memory to the GPU before calculation, while the copyout clause allocates an empty buffer and copies it back to the host memory after leaving. This can reduce the amount of data that needs to be transferred between different parts of the system.

[0038] In order to improve the operating efficiency of parallel computing and avoid problems such as memory leakage or excessive memory usage, the present invention also optimizes the data allocation rules and uses dynamic memory allocation of data to save memory space while reducing the amount of memory required for temporary variables between computing areas.

[0039] In the process of calculating the visual factor, the traditional RGM model uses a line scanning algorithm for area filling and subsequent projection area calculation, but this method is computationally inefficient and difficult to adapt to GPU acceleration. Therefore, the embodiment of the present invention uses a triangle filling algorithm based on barycentric coordinates (TFBCs) to replace the traditional line scanning algorithm for area filling and projection area calculation. Specifically, the Cartesian coordinates of any point in three-dimensional space are converted to barycentric coordinates, and the barycentric coordinates are used to determine whether the point is located inside a certain triangular face element. In the polygon filling algorithm, any point P can be expressed as follows using the coordinates of the three vertices A, B, and C of the face element: ; in, and represents the centroid coordinates of point P; when and If both are non-negative numbers, then point P is inside the bin; otherwise, point P is outside the bin.

[0040] As for the center of gravity coordinates and It can be calculated by the following formula: ; in, , and Represents vectors , and The component in the x direction, , and Represents vectors , and Component in the y direction.

[0041] By filling in the area and calculating the projected area, the visible factor of each face element can be determined. For the calculated data, the embodiment of the present invention adopts a hash-based data structure for storage. By replacing the traditional array storage structure, the data query efficiency is improved from O(n) to O(1), which greatly improves the data access speed and processing efficiency.

[0042] S4: After the visibility factor is calculated, the radiation flux density of the surface element is iteratively calculated using the successive over-relaxation method (SOR). The successive over-relaxation method (SOR) introduces a weighted average method into the Gauss-Seidel method used in the traditional radiation flux density iterative calculation process. The iterative calculation formula of the radiation flux density of the surface element is: ; in, represents the relaxation factor, represents the k+1th iteration radiation flux density of the i-th surface element, represents the k-th iteration radiation flux density of the i-th surface element, represents the k-th iteration radiation flux density of the j-th surface element, represents the reflectivity or transmittance of the jth face element, N represents the total number of face elements, represents the visibility factor from the jth bin to the ith bin, It represents the flux density of spontaneous radiation of the ith surface element under sunlight. The value range is 0 to 2, and the relaxation factor The value of can be optimized and adjusted. In the process of solving the radiation flux density, the optimal relaxation factor is searched within the range of relaxation factors with a step size of 0.2 to ensure the fastest iteration process and guarantee the calculation accuracy.

[0043] S5: After calculating the radiation flux density of each face element, the bidirectional reflectivity BRF of the face element at different observation angles can be further calculated. The calculation formula of BRF is as follows: ; in, represents the observation angle, Indicated in Bidirectional reflectivity at angles, represents the normal direction of the i-th face element, From the observation point of view The area of ​​the i-th face element seen.

[0044] The above three-dimensional radiosity calculation method of high-resolution remote sensing scenes can be used to optimize the traditional RGM model and construct an improved GPU-based three-dimensional radiosity graphics model (RGM PS), the model can be applied to complex heterogeneous canopy simulation. Using this model for simulation can not only improve the speed of visual factor calculation, but also ensure the model processing accuracy. In addition, due to the significant improvement in the speed of visual factor calculation, RGM PS The model significantly increases the complexity of simulated scenes from tens of thousands of faces in the original model to 14 million faces, enabling it to meet the simulation requirements of large-scale, high-resolution remote sensing scenes.

[0045] The three-dimensional radiometric calculation method of the high-resolution remote sensing scene provided in the embodiment of the present invention has been experimentally verified. Specifically, the RGM is constructed by the above method. PS The model was validated by real single tree experiments and synthetic heterogeneous canopy experiments. PS The correlation (R²) between the model and the original RGM model exceeded 0.94, the root mean square error (RMSE) was less than 0.0038, and the cross-validation with the DART model showed high consistency (R² as high as 0.98). The results of the airborne multi-angle measurement evaluation showed that the RGM PS The model has good accuracy in BRF simulation, with RMSE of 0.0031 and 0.0340 in the red and near-infrared bands, respectively.

[0046] The specific verification experiments and experimental results are as follows: Experimental goal 1: Verify RGM PS The model is progressive in computational speed and accuracy.

[0047] The experiment simulates the target selection of a real single tree scene and compares the original RGM model with the RGM PS The performance efficiency of the model. The experiments were conducted using a Dell laptop equipped with an Intel1 Core2 i7-12700H CPU @2.30 GHz, 32G RAM, and an NVIDIA GeForce RTX3060 GPU. The BRF simulation time of the original RGM model is 2 hours, 53 minutes, and 18 seconds (10398 seconds), while the RGM PS The model took only 2 minutes and 32 seconds (152 seconds), achieving a speedup of about 68.41 times.

[0048] Experimental goal 2: Verify the effectiveness of the acceleration strategy under the OpenACC parallel framework.

[0049] To verify the effectiveness of different improvement strategies, ablation experiments were conducted to determine the performance characteristics: The speedup ratio of parallel loop instructions is 4.32; the speedup ratio of data layout optimization is 4.30; the speedup ratio of minimizing data transfer is 3.57; the speedup ratio of data allocation is 1.33; the speedup ratio of the improved polygon filling algorithm is 53.68; and the speedup ratio of solving the radiation flux density equation of the new face element is 1.21. The polygon filling algorithm and the hash map-based data structure have a significant impact on performance efficiency and show significant acceleration effects. In contrast, parallel loop instructions and optimized data layout strategies, which are common ported operations in the OpenACC framework, show suboptimal acceleration effects, achieving an improvement of about 4.3 times. The other three strategies show relatively moderate parallel computing performance, but together they are RGM PS It contributes to an overall computational speed-up of about 6 times. Experiments were conducted on different facet sizes ranging from 1,000 to 100,000 to investigate the efficiency of the acceleration strategy. The results confirmed that the overall speed-up ratio remained stable throughout the experiment, ranging from 60.2 to 71.1 times.

[0050] Experimental goal 3: Cross-validation with the existing three-dimensional radiation transfer model DART Utilizing RGM PS The model and DART model were simulated to obtain Figure 3 The reflectivity results shown are Figure 3 a, b, c and d are the principal plane reflectance results. The principal plane is the plane containing the sensor optical axis and the nadir point, which can be understood as the plane containing the sun, sensor and ground. Figure 3 In the figure, e, f, g and h are the cross-principal plane reflectance results. The cross-principal plane is a plane perpendicular to the principal plane and contains the nadir point. As can be seen from the figure, there is a significant similarity between the two models. The overall curve shape and trend of the reflectance BRF changing with the observation azimuth are comparable. So we can get RGM PS The conclusion that the simulated reflectivity value obtained by the DART model is highly consistent verifies that the method of the embodiment of the present invention greatly improves the calculation speed without reducing the accuracy of the reflectivity simulation calculation.

[0051] Experimental goal four: Conduct comparative experiments with ground-truth airborne measurements.

[0052] like Figure 4 As shown, RGM PS The simulation results of DART and 4SAIL models are compared with the real scene airborne measurement results. For the red band and near infrared band, RGM PS The R of the model for canopy BRF 2The values ​​are 0.90 and 0.83, and the RMSE values ​​are 0.0031 and 0.0340, respectively. The results are highly consistent with the airborne measurement results of real scenes, proving that the RGMPS model has high spatial simulation accuracy and can be used to simulate large-scale, high-resolution remote sensing scenes such as vegetation, forests and urban scenes.

[0053] In short, the above description is only a preferred embodiment of this specification and is not intended to limit the protection scope of this specification. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this specification shall be included in the protection scope of this specification.

[0054] The systems, devices, modules or units described in one or more of the above embodiments may be implemented by a computer chip or entity, or by a product having a certain function. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0055] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0056] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0057] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for calculating three-dimensional radiometry of a high-resolution remote sensing scene, characterized in that: include: By defining vertices in three-dimensional space to generate a triangular facet mesh, a three-dimensional scene is created; Calculate the flux density of spontaneous radiation of the surface element under sunlight based on the optical parameters of the surface element; Calculating the visual factor through the OpenACC parallel framework, including: transferring the face data used to calculate the visual factor from the CPU to the GPU, allocating the calculation tasks of the visual factor loop iteration between different facets to multiple threads of the GPU, and accelerating the visual factor calculation; The radiation flux density of the surface element is iteratively calculated using the successive super-relaxation method. The iterative calculation formula is: ； in, represents the relaxation factor, represents the k+1th iteration radiation flux density of the i-th surface element, represents the k-th iteration radiation flux density of the i-th surface element, represents the k-th iteration radiation flux density of the j-th surface element, represents the reflectivity or transmittance of the jth face element, N represents the total number of face elements, represents the visibility factor from the jth bin to the ith bin, represents the flux density of spontaneous radiation of the i-th surface element under sunlight; The bidirectional reflectivity BRF at different observation angles is calculated based on the radiation flux density of each surface element. The calculation formula is: ； in, represents the observation angle, Indicated in Bidirectional reflectivity at angles, represents the normal direction of the i-th face element, From the observation point of view The area of ​​the i-th face element seen, Represents the radiation flux density of the i-th surface element.

2. The three-dimensional radiometry calculation method of high-resolution remote sensing scene according to claim 1, characterized in that: The calculation formula of the flux density of spontaneous radiation of the surface element under sunlight is: ； in, represents the area of ​​the ith facet, Represents any point on the i-th face The flux density of spontaneous emission.

3. The three-dimensional radiometry calculation method of high-resolution remote sensing scene according to claim 1, characterized in that: The calculation formula of the visual factor is: ； in, Express differential, Express differential, express and Connection between The angle of the surface normal, express and Connection between The angle of the surface normal, express and The Euclidean distance between .

4. The three-dimensional radiometry calculation method of high-resolution remote sensing scene according to claim 3, characterized in that: Under the OpenACC parallel framework, at least one strategy among parallel loop instructions, optimized data layout, minimized data transfer and dynamic memory allocation is adopted to accelerate the field of view factor calculation.

5. The three-dimensional radiometry calculation method of high-resolution remote sensing scene according to claim 4, characterized in that: The polygon filling algorithm is used to fill the area of ​​the surface element, and the Cartesian coordinates of any point in the three-dimensional space are converted into barycentric coordinates. The barycentric coordinates are used to determine whether the point is located in a certain surface element.

6. The method for calculating three-dimensional radiometry of a high-resolution remote sensing scene according to claim 5, characterized in that: In the polygon filling algorithm, the coordinates of the three vertices A, B and C of the face element are used to represent any point P as: ； in, and represents the centroid coordinates of point P; when and If both are non-negative numbers, then point P is inside the bin; otherwise, point P is outside the bin.

7. The method for calculating three-dimensional radiometry of a high-resolution remote sensing scene according to claim 6, characterized in that: Barycentric coordinates and The calculation formula is: ； in, , and Represents vectors , and The component in the x direction, , and Represents vectors , and Component in the y direction.

8. The method for calculating three-dimensional radiometry of a high-resolution remote sensing scene according to claim 1, characterized in that: During the visual factor calculation process, the storage structure in the GPU is a hash-based data structure.

9. The method for calculating three-dimensional radiometry of a high-resolution remote sensing scene according to claim 1, characterized in that: The range of the relaxation factor is from 0 to 2. During the solution of the radiation flux density, the optimal relaxation factor is searched within the range of the relaxation factor with a step size of 0.

2.

10. An application of the three-dimensional radiometry calculation method for high-resolution remote sensing scenes according to any one of claims 1 to 9, characterized in that: A three-dimensional radiosity graphics model based on a GPU is constructed using the three-dimensional radiosity calculation method for a high-resolution remote sensing scene as described in any one of claims 1 to 9, and the three-dimensional radiosity graphics model is used for heterogeneous canopy simulation.

Images