Rapid deterministic channel modeling method based on improved bounding box ray tracing

Abstract

The invention relates to the technical field of ray tracing, and provides a fast deterministic channel modeling method based on improved bounding box ray tracing, which comprises the following steps: acquiring a scene containing a plurality of buildings, and determining a surface area heuristic value of a potential segmentation point in the scene; the potential segmentation point is a point used for dividing the space represented by the current node into two subspaces in the process of constructing the hierarchical bounding box tree; the node is a unit in the hierarchical bounding box tree structure; acquiring a set of position surface area heuristic values of all potential segmentation points, and finding out a minimum surface area heuristic value in the set as an optimization problem for modeling; and obtaining a preset threshold value, and when the number of patches in the space represented by the nodes is greater than the threshold value, solving an optimization problem, determining a segmentation plane, and constructing a hierarchical bounding box tree structure by a particle swarm optimization-based traversal segmentation point optimization method.

Description

Technical Field

[0001] The invention relates to the technical field of ray tracing, and in particular to a fast deterministic channel modeling method based on improved bounding box ray tracing. Background Art

[0002] In complex and ever-changing scenarios such as emergency rescue, quickly and accurately establishing wireless channel models is the key to ensuring stable communication of unmanned aerial vehicles (UAVs). This demand has driven the research on efficient channel modeling methods. Ray Tracing (RT), as a deterministic channel modeling technology, is based on geometric optics (GO), geometric diffraction theory (GTD) and consistent diffraction theory (UTD). It calculates channel parameters by simulating the propagation of light in a three-dimensional environment, thereby providing a high-precision channel model. However, when dealing with large-scale and complex scenarios, the ray tracing method requires high-resolution ray emission and detailed path tracing, resulting in huge computational complexity and slow modeling speed, making it difficult to meet the real-time requirements in emergency communications.

[0003] In the prior art, in order to accelerate the ray tracing process, the bounding volume hierarchy (BVH) algorithm is widely used. BVH effectively reduces unnecessary intersection detection between rays and objects in the scene by establishing a hierarchical spatial division of the scene, thereby significantly improving the efficiency of ray tracing. However, the BVH construction process itself, especially the selection of segmentation planes, is still a computationally intensive task. Traditional segmentation plane selection methods, such as the surface area heuristic (SAH), can evaluate the probability of intersection between rays and objects based on the surface area of ​​geometric bodies. However, when faced with large-scale scenes, the computational cost of traversing all candidate segmentation points to find the best segmentation plane is extremely high, resulting in a longer construction time for the BVH tree, which is difficult to meet the needs of rapid modeling of dynamically changing scenes.

[0004] In view of the above background, how to quickly and effectively build a high-quality BVH tree to accelerate the ray tracing and channel modeling process has become an important issue that needs to be urgently solved in the industry. Summary of the invention

[0005] The present invention provides a fast deterministic channel modeling method based on improved bounding box ray tracing, which is used to solve the defects of the prior art that the hierarchical bounding box BVH tree construction time is long and it is difficult to meet the rapid modeling needs in scenarios with high real-time requirements such as emergency rescue, thereby accelerating the construction process of the BVH tree, effectively reducing the modeling time and improving the efficiency of ray tracing.

[0006] The present invention provides a fast deterministic channel modeling method based on improved bounding box ray tracing, comprising the following steps: Obtain a scene containing multiple buildings, and determine a surface area heuristic value of a potential segmentation point in the scene; the potential segmentation point is a point used to divide the space represented by the current node into two subspaces during the process of constructing a hierarchical bounding box tree; the node is a unit in the hierarchical bounding box tree structure; Obtaining a set of position surface area heuristic values ​​of all potential segmentation points, and modeling the method of finding a minimum surface area heuristic value in the set as an optimization problem; A preset threshold value is obtained. When the number of facets in the space represented by the node is greater than the threshold value, the optimization problem is solved based on a traversal segmentation point optimization method of particle swarm optimization, the segmentation plane is determined, and a hierarchical bounding box tree structure is constructed.

[0007] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the step of obtaining a scene containing multiple buildings and determining a surface area heuristic value of a potential segmentation point in the scene specifically includes: obtaining a scene containing multiple buildings, traversing each face in the scene and determining its position relative to the potential segmentation point; for each potential segmentation point, determining the time complexity of constructing a hierarchical bounding box tree using a surface area heuristic algorithm according to the number of faces contained in the potential segmentation point, and using the time complexity as the surface area heuristic value of the segmentation point.

[0008] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the traversal segmentation point optimization method based on particle swarm optimization solves the optimization problem, specifically including: obtaining the current scene face information and the bounding box volume list including the face as the input of the particle swarm optimization algorithm; determining the particle population size and the number of iterations according to the scene size and the number of facets; and traversing the potential segmentation points based on the particle swarm optimization algorithm according to the input, the particle population size and the number of iterations to solve and determine the optimal solution.

[0009] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the obtaining of current scene patch information and a list of bounding box volumes including the patches specifically comprises: sorting n patches in the current scene in sequence according to spatial positions in the direction of a certain coordinate axis to obtain an ordered queue of all patches; based on the ordered queue, obtaining a defined patch information list, and determining an information set of all patches according to the patch information list; based on the ordered queue, defining a left bounding box volume including the first i patches from left to right, and determining a left sub-bounding box volume set according to the left bounding box volume; defining a right bounding box volume including the first i patches from right to left, and determining a right sub-bounding box volume set according to the right bounding box volume.

[0010] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the potential segmentation points are traversed based on the particle swarm optimization algorithm according to the input, the particle population size and the number of iterations to solve and determine the optimal solution, which specifically includes: initializing the population: initializing a population containing k particles, and randomly initializing the particle positions in an area where the surface area heuristic value is greater than a preset threshold; evaluating all particles, setting the individual optimal position of each particle and the population optimal position; evaluating particles: according to the position of each particle, obtaining the index of the bounding box volume list corresponding to it, and calculating the fitness value of each particle based on the fitness function according to the left and right bounding box volumes corresponding to the index and the number of facets inside the bounding box; updating the particle state: for each particle, if the current position fitness value is better than its individual optimal fitness, then update its individual optimal position, if the current position fitness value is better than its population optimal fitness, then update its population optimal position; repeat the steps of evaluating particles and updating particle states until the optimal segmentation plane is determined after the iterative convergence condition is met.

[0011] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the construction of a hierarchical bounding box tree structure specifically includes: parsing the positions of all facets in the scene; creating a current node, and enclosing all facets of the current node with a bounding box; determining the number of facets in the current node, and when the number of facets in the current node is greater than a preset number threshold, finding the best splitting plane through a particle swarm optimization algorithm to divide the current node, and outputting a left and right node facet set as the left and right child nodes of the current node; continuing to recursively divide the divided child nodes as the current node until the number of facets in the current node is no greater than the preset number threshold, stopping the recursion, marking the current node as a leaf node and returning it; and outputting the hierarchical bounding box tree structure after the recursion ends.

[0012] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, after modeling with finding the minimum surface area heuristic value in the set as an optimization problem, the method also includes: obtaining a preset threshold value, and when the number of facets in the space represented by the node is not greater than the threshold value, using a surface area heuristic traversal method to find the optimal segmentation plane.

[0013] The present invention also provides a fast deterministic channel modeling device based on improved bounding box ray tracing, comprising the following modules: A surface area heuristic value acquisition module is used to acquire a scene containing multiple buildings and determine a surface area heuristic value of a potential segmentation point in the scene; the potential segmentation point is a point used to divide the space represented by the current node into two subspaces during the process of constructing a hierarchical bounding box tree; the node is a unit in the hierarchical bounding box tree structure; An optimization problem modeling module is used to obtain a set of position surface area heuristic values ​​of all potential segmentation points, and find a minimum surface area heuristic value in the set as an optimization problem for modeling; The hierarchical bounding box tree construction module is used to obtain a preset threshold value. When the number of facets in the space represented by the node is greater than the threshold value, the optimization problem is solved based on the particle swarm optimization traversal segmentation point optimization method, the segmentation plane is determined, and the hierarchical bounding box tree structure is constructed.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the fast deterministic channel modeling method based on improved bounding box ray tracing as described in any one of the above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described fast deterministic channel modeling methods based on improved bounding box ray tracing.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the fast deterministic channel modeling method based on improved bounding box ray tracing as described in any one of the above is implemented.

[0017] The present invention provides a fast deterministic channel modeling method based on improved bounding box ray tracing. By constructing a hierarchical bounding box tree structure, the number of scene elements that need intersection testing during ray tracing can be greatly reduced. The hierarchical bounding box tree recursively divides the scene into smaller parts and only performs detailed intersection testing on the parts that may intersect with the rays, thereby significantly improving the efficiency of ray tracing. By minimizing the surface area heuristic value, the scene can be divided more effectively so that each node more tightly surrounds the geometry inside it, thereby reducing unnecessary intersection testing. By introducing a particle swarm optimization (PSO) algorithm to accelerate the construction process of a hierarchical bounding box (BVH) tree, a large amount of calculation required for selecting the best segmentation plane in traditional methods is effectively reduced, and the construction time of the BVH tree is significantly shortened. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention.

[0020] Figure 2a It is a schematic diagram of layer-by-layer segmentation of the BVH bounding box provided by the present invention.

[0021] Figure 2b It is a schematic diagram of forming a BVH tree based on the segmented BVH bounding box provided by the present invention.

[0022] Figure 3 It is a schematic diagram of the process of constructing a hierarchical bounding box tree based on particle swarm optimization provided by the present invention.

[0023] Figure 4a It is a side view of the ray tracing simulation result in the scene provided by the present invention.

[0024] Figure 4b It is a top view of the ray tracing simulation result in the scene provided by the present invention.

[0025] Figure 5a This is one of the comparison graphs of the time required for running the direct traversal ray tracing algorithm provided by the present invention and the other three acceleration algorithms.

[0026] Figure 5b This is one of the running time comparison diagrams of the SAH_PSO algorithm provided by the present invention, the conventional BVH-SAH algorithm, and the kdTree algorithm.

[0027] Figure 6a This is the second comparison chart of the time required for running the direct traversal ray tracing algorithm provided by the present invention and the other three acceleration algorithms.

[0028] Figure 6b This is the second running time comparison chart of the SAH-PSO algorithm provided by the present invention, the conventional BVH-SAH algorithm, and the kdTree algorithm.

[0029] Figure 7 It is a schematic diagram comparing the time consumption of searching for the best segmentation point in a single time of the SAH-PSO algorithm provided by the present invention and its benchmark algorithm BVH-SAH.

[0030] Figure 8It is a structural schematic diagram of a fast deterministic channel modeling device based on improved bounding box ray tracing provided by the present invention.

[0031] Fig. 9 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0033] In the process of emergency rescue, it is necessary to establish a stable and accurate wireless channel model for complex and changing scene environments, provide stable communication connections for unmanned aerial vehicles (UAVs) at different locations, guide the UAVs to carry out subsequent location deployment according to the signal strength and line-of-sight link conditions at different locations, and use the base stations on board to quickly build a network to provide wireless coverage for ground users. Ray Tracing (RT) is a deterministic channel modeling method based on the theory of geometrical optics (GO), geometrical diffraction theory (GTD) and uniform diffraction theory (UTD). The transmitter emits a large number of rays in all directions at a certain angle interval (i.e., resolution), studies the direct, reflected, diffracted, and refracted paths of the rays in a three-dimensional environment, and uses electromagnetic field theory to calculate the received power, delay spread, arrival angle, frequency offset and other parameters of each ray. Ray tracing has solved the defects of statistical channel modeling methods due to their reliance on statistical empirical formulas and low coupling with the actual environment to a certain extent. However, in pursuit of modeling accuracy, the ray tracing method needs to emit high-resolution ray clusters at the transmitting antenna and track the path of each ray and the channel parameters at different receiving ends, which results in a huge amount of computation and slow channel modeling. Special communication scenarios such as emergency command communications have extremely high requirements for the time to establish the channel model of the user's environment, and channel modeling usually needs to be completed within minutes or even seconds. In addition, the time-varying nature of the communication environment also requires the channel modeling method to have a fast time response. Therefore, how to accelerate the ray tracing method and quickly and efficiently establish the channel model is one of the challenging problems that the channel modeling method based on ray tracing needs to solve.

[0034] At present, some work has studied the acceleration algorithm of ray tracing. The bounding volume hierarchy (BVH) algorithm is a typical representative method, which achieves the purpose of acceleration by improving the efficiency of collision detection between rays and objects. BVH usually uses some geometric bodies with relatively simple geometric features to wrap objects with complex structures in the scene. When detecting whether the ray intersects with the objects in the scene, the ray and the bounding box containing the scene object are first intersected: if they do not intersect, it means that the scene object currently tested cannot intersect with the ray; if they intersect, the ray is then tested for intersection with a smaller bounding box in the bounding box until a leaf node is found, that is, the object intersected with the ray. Although BVH can establish a hierarchical bounding box tree for the scene and provide a good acceleration effect in the ray traversal process, its tree building process is computationally intensive and takes a long time to build. In command and communication scenarios such as emergency rescue, the acceleration effect is not good in the face of the dynamic nature of the scene. Therefore, it is necessary to propose an algorithm for quickly building a hierarchical bounding box tree to accelerate the tree building process. Existing research mainly uses parallel construction methods to accelerate ray tracing in parallel. A high-performance thread is used to calculate the upper large nodes until the calculation amount of each node is small enough, and then they are allocated to the structure of the BVH tree. However, the parallel method ultimately relies on the acceleration technology of specific hardware, and the data transmission between the GPU and the host memory will introduce additional transmission overhead, which will become a performance bottleneck when the data volume is large. Therefore, it is of great research value to accelerate ray tracing at the algorithm level to assist in channel modeling. When the BVH algorithm selects the node position, the surface area heuristic (SAH) is used to select the best splitting plane. The idea of ​​SAH is to measure the probability of the ray intersecting with the geometric body in the scene by a quantity related to the surface area of ​​the geometric body. After the ray enters the scene, it is possible to intersect with any object in the scene. When the ray is evenly distributed in space, the probability of intersecting with the scene object with a larger surface area must be greater than that of the scene object with a relatively small surface area.

[0035] In the process of building a tree and determining the node position, the BVH algorithm needs to use SAH to select the best splitting plane. In order to solve the problem that the traditional BVH algorithm is slow in selecting node positions and takes a long time to build a tree, the present invention proposes an optimal bounding box node position decision scheme based on a heuristic algorithm. The original SAH strategy of traversing all candidate split points to select the best splitting plane is improved, thereby accelerating the tree building process and making the algorithm more suitable for dynamically changing scenarios. Specifically, the present invention introduces the particle swarm optimization (PSO) algorithm to use the index corresponding to the patch as the particle position, and the time cost of accessing the node under the current partitioning strategy as the fitness function. Through the iteration of particles, the best position is selected, that is, the node with the smallest traversal cost, thereby establishing a high-quality hierarchical bounding box tree, which effectively accelerates the ray tracing process.

[0036] The present invention mainly solves the problem of fast deterministic channel modeling through ray tracing in emergency communication scenarios. In emergency scenarios, the scene may change at any time, and each change in the scene requires the hierarchical bounding box tree to be re-established. The traditional SAH method takes a long time to establish the tree, making it difficult to provide fast services and meet the real-time requirements of emergency rescue. The present invention proposes a fast bounding box tree building method based on particle swarms, which uses the position of particles as candidate segmentation planes, evaluates the pros and cons of particle positions by comparing the time cost required to traverse each segmentation plane, and quickly finds the best segmentation plane, thereby establishing a high-quality BVH tree to accelerate the channel modeling process.

[0037] Combine the following Figure 1-Figure 9 Embodiments of the present invention are described.

[0038] Figure 1 is one of the flow charts of the fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, such as Figure 1 As shown, the method includes the following: S110 , obtaining a scene containing multiple buildings, and determining surface area heuristic values ​​of potential segmentation points in the scene.

[0039] A potential split point is a point used to divide the space represented by the current node into two subspaces during the process of constructing a hierarchical bounding box tree; a node is a unit in the hierarchical bounding box tree structure.

[0040] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, a scene containing multiple buildings is obtained, and a surface area heuristic value of a potential segmentation point in the scene is determined, specifically comprising: obtaining a scene containing multiple buildings, traversing each face in the scene and determining its position relative to the potential segmentation point; for each potential segmentation point, determining the time complexity of constructing a hierarchical bounding box tree using a surface area heuristic algorithm according to the number of faces contained in the potential segmentation point, and using the time complexity as the surface area heuristic value of the segmentation point.

[0041] Specifically, the present invention assumes that a scene containing multiple buildings is to be considered. According to the faces of all buildings in the scene, a high-quality BVH tree is constructed in a short time, so that the subsequent process rays can quickly traverse all the faces in the scene. This process requires the use of the SAH algorithm to determine the segmentation strategy to minimize the traversal cost. The three premise assumptions of the SAH algorithm are as follows: (1) The rays in space are uniformly distributed, and their propagation in space will not be blocked or change their propagation speed.

[0042] (2) If the number of all faces in the scene is n, then each ray detects the scene n times.

[0043] (3) Assume that the time complexity of intersection detection between a single object and a ray is equal and is , then for the intersection detection of n faces, the total time complexity is

[0044] Based on the above three assumptions, a specific description of the traversal cost can be obtained: for an arbitrarily distributed ray, the probability P of the ray passing through a sub-region of a spatial geometry is: (1)

[0045] in Representing Geometry The surface area of Representation scene As shown in Figure 2 (b), in the tree structure, A is the root node of the tree, B, C, and D are child nodes, and E, F, G, H, and I are all leaf nodes. Assume that the computational cost of a ray visiting a non-leaf node is , the computational cost of a ray visiting a leaf node is , so the following SAH cost formula can be given: (2)

[0046] (3)

[0047] When a split point is given When , the SAH value at the split point is calculated by formula (2), which is the sum of the computational cost of the node and the SAH values ​​of its two child nodes. and Represent the left and right child nodes respectively. Formula (3) represents the computational cost of accessing the entire tree, which is the sum of the computational costs of accessing all leaf nodes and non-leaf nodes. From formula (2), it can be seen that if you want to accurately calculate the optimal segmentation point of a node, that is, the position of the minimum SAH value, you must calculate the SAH values ​​of all its child nodes through subsequent traversal. However, the subsequent traversal needs to calculate all possible division situations. If the number of objects in the scene increases, the number of feasible segmentation points will increase exponentially. If the scene is slightly more complicated, it will cause the optimization strategy to generate a huge amount of calculation. Therefore, we can only make a second choice and find the local optimal solution through estimation, using the number of patches corresponding to the node to replace the node access cost. And the SAH value of the child node, that is, using the local greedy SAH instead of the global optimization strategy, so that each division of the node is locally optimal, therefore, formula (2) is modified as follows: (4)

[0048] in, and Respectively represent the number of facets contained in the left and right subnode spaces of the split point, Representing Geometry The surface area of Representing Geometry surface area.

[0049] S120, obtaining a set of position surface area heuristic values ​​of all potential segmentation points, and finding the minimum surface area heuristic value in the set as an optimization problem for modeling.

[0050] Specifically, an optimization problem is established.

[0051] The most time-consuming step in constructing a BVH tree based on the SAH algorithm is to determine the position of each segmentation point, which requires traversing the position of each face in the scene to compare the minimum SAH value and determine the best segmentation plane based on the segmentation point with the minimum SAH value. Without loss of generality, assuming that the parent node space containing N faces is divided into two child node spaces of roughly equal size and containing N / 2 faces, the time cost required is , and the recursion continues until N=1, that is, each leaf node contains only a single face. The cost of building a whole tree for N faces is: (5)

[0052] To simplify further, we define the time cost of calculating the SAH value at a single patch as = 1, then equation (5) becomes: (6)

[0053] Therefore, the time complexity of constructing a hierarchical bounding box tree using the SAH algorithm is .

[0054] make is the set of SAH values ​​of all feasible segmentation point positions, and all segmentation point positions are within the scene limit. Indicates that the range of segmentation points that can be selected is limited to the middle of the leftmost and rightmost positions of the scene patch. The goal is to find the best segmentation point position , which satisfies the minimum traversal cost SAH value of establishing a sub-node by splitting the scene from this position. The optimization problem can be modeled as: P: st,

[0055] C1: (7)

[0056] C2: C3: Among them, C1 means that all candidate segmentation points are within the scene boundary limit, and C2 means that the SAH value of all feasible segmentation points should be less than a cost threshold , the segmentation point position with too large SAH value is not considered, C3 means that the number of facets contained in the leaf node should be less than the given number threshold , when the number of child nodes is less than , return to the leaf node and no longer traverse and optimize.

[0057] S130, obtaining a preset threshold value, and when the number of facets in the space represented by the node is greater than the threshold value, solving the optimization problem based on the particle swarm optimization traversal segmentation point optimization method, determining the segmentation plane, and constructing a hierarchical bounding box tree structure.

[0058] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the optimization problem is solved by a traversal segmentation point optimization method based on particle swarm optimization, which specifically includes: obtaining the current scene patch information and a list of bounding box volumes including the patches as inputs to the particle swarm optimization algorithm; determining the particle population size and the number of iterations according to the scene size and the number of patches; and traversing potential segmentation points based on the particle swarm optimization algorithm according to the input, the particle population size and the number of iterations to determine the optimal solution.

[0059] Specifically, formula (7) is a nonlinear optimization problem with a large range of variable values. Although the SAH traversal can be used to solve the problem, it can obtain the optimal solution and find the best segmentation strategy, but it is an exhaustive search method with large amount of calculation and long calculation time. For emergency communication scenarios, it is crucial to quickly establish a BVH tree to assist ray tracing for channel modeling. Therefore, the present invention proposes a traversal segmentation point optimization method based on particle swarm optimization (PSO). By simulating swarm intelligence, it avoids evaluating every feasible solution in the entire search space, only evaluates partial solutions, and gradually approaches the optimal solution through iteration, so that the amount of calculation is significantly reduced. The proposed algorithm is divided into two steps: (1) Preparation process: In this algorithm, the particle swarm algorithm cannot be run directly, and it is necessary to provide the prior information required by its fitness function. Specifically, the present invention first calculates the volume list of the bounding box containing the current scene patch and uses it as the input of the PSO algorithm.

[0060] (2) Optimization process: According to the size of the scene and the number of faces, an appropriate population size and number of iterations are given to quickly find the optimal solution, determine the segmentation plane, and construct a BVH tree.

[0061] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, the current scene face information and the bounding box volume list including the facets are obtained, which specifically includes: sorting the n facets in the current scene in sequence according to the spatial position in the direction of a certain coordinate axis to obtain an ordered queue of all facets; based on the ordered queue, obtaining a defined facet information list, and determining an information set of all facets according to the facet information list; based on the ordered queue, defining a left bounding box volume including the first i facets from left to right, and determining a left sub-bounding box volume set according to the left bounding box volume; defining a right bounding box volume including the first i facets from right to left, and determining a right sub-bounding box volume set according to the right bounding box volume.

[0062] Specifically, prior information is calculated.

[0063] In the algorithm prior information preparation stage, it is first necessary to calculate the input parameters required by the PSO fitness function. In the present invention, the input parameters include the number of facets and the list of bounding box volumes containing the facets. Specifically, there are two steps: (a) First, sort the n faces in the scene according to their spatial positions along a certain coordinate axis (usually the x-axis) to obtain an ordered queue of all faces. ; Definition list Indicates The minimum value of the three-dimensional coordinates of all vertices of the face and maximum value ; Then, calculate the information set of all patches .

[0064] (b) Definition Indicates that counting from left to right includes the previous The volume of the left bounding box within the face is required to obtain the volume set of the left sub-bounding box ;definition Indicates that counting from right to left includes the previous The right bounding box volume within the face is required to obtain the right sub-bounding box volume set .

[0065] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, a particle swarm optimization algorithm is used to traverse potential segmentation points for solution according to input, particle population size and iteration times to determine the optimal solution, specifically including: initializing the population: initializing a population containing k particles, and randomly initializing the particle positions in an area where the surface area heuristic value is greater than a preset threshold; evaluating all particles, setting the individual optimal position of each particle and the population optimal position; evaluating particles: according to the position of each particle, obtaining the index of the bounding box volume list corresponding to it, and calculating the fitness value of each particle based on the fitness function according to the left and right bounding box volumes corresponding to the index and the number of facets inside the bounding box; updating the particle state: for each particle, if the current position fitness value is better than its individual optimal fitness, then update its individual optimal position, if the current position fitness value is better than its population optimal fitness, then update its population optimal position; repeat the steps of evaluating particles and updating particle states until the optimal segmentation plane is determined after the iterative convergence condition is met.

[0066] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, after modeling with finding the minimum surface area heuristic value in a set as an optimization problem, a preset threshold value is obtained, and when the number of facets in the space represented by the node is not greater than the threshold value, a surface area heuristic traversal method is used to find the optimal segmentation plane.

[0067] Specifically, a single segmentation plane optimization is performed based on PSO.

[0068] In order to minimize the BVH tree building time, the present invention comprehensively considers the impact of the PSO algorithm on the process of building a hierarchical bounding box tree. According to formula (6), the time complexity of directly performing SAH traversal optimization on the scene to build a BVH tree is: , the time complexity of searching for the best splitting plane is , the time complexity of using PSO to find the best splitting plane is , and Represent the particle population size and the number of iterations respectively. When the number of facets in a node is large, it can be seen that the PSO algorithm has a great advantage and can greatly reduce the time complexity. As the recursion proceeds, the number of facets in the node gradually decreases. < If PSO is called again for optimization, it will increase the time complexity of the algorithm and take longer than directly traversing the optimization. Therefore, a reasonable threshold value needs to be set. , when the number of patches in the node is not greater than the threshold When , the PSO algorithm is no longer called to find the best splitting plane, and the SAH traversal method is used instead.

[0069] The specific process of PSO to find the best splitting plane is as follows: (a) Initialization phase: The initial population includes Particles In order to reduce the iteration time of the algorithm, instead of randomly giving particle positions in the entire search space, the particle positions are randomly initialized in the area with a relatively large SAH value, which can make the algorithm converge faster. Finally, all particles are evaluated and the optimal position of each particle is set accordingly. and the optimal position of the entire population .

[0070] (b) Evaluation phase: Based on the position of each particle, the index of its corresponding volume list can be obtained. According to the volume of the left and right bounding boxes corresponding to the index and the number of faces inside the bounding box, the fitness value of each particle can be calculated. Define the fitness function: (8)

[0071] in, is the number of patches in the node, is the index of the current patch. Equation (8) is rewritten from equation (4). The fitness function corresponds to the SAH value of the cutting plane. Therefore, the process of PSO finding the optimal solution just fits the process of the algorithm finding the minimum SAH value.

[0072] (c) Update phase: For each particle , if the current position adapts to the value Better than its individual optimal fitness , then update the individual optimal position , if the current position adapts to the value Better than the optimal fitness of its population , then update the optimal position of the population Finally, for each particle Update the speed according to the following formula : (9)

[0073] in, represents the number of iterations, is the inertia weight, which is used to balance the capabilities of local search and global search. The larger the value, the stronger the global optimization ability is, but the local optimization ability is weaker. and are learning factors, representing the learning abilities of individuals and groups, and is a random number used to introduce randomness to prevent the algorithm from falling into a local optimal solution. , since the position of the particle represents the index of the list, its value can only be a positive integer in the search space. Therefore, compared with the conventional PSO position update formula, the PSO position update formula in the present invention requires an additional approximate rounding operation: (10)

[0074] The details of the algorithm for finding the best segmentation plane using PSO are shown in Table 1:

[0075] Table 1

[0076] According to a fast deterministic channel modeling method based on improved bounding box ray tracing provided by the present invention, a hierarchical bounding box tree structure is constructed, which specifically includes: parsing the positions of all facets in the scene; creating a current node, and enclosing all facets of the current node with a bounding box; determining the number of facets in the current node, when the number of facets in the current node is greater than a preset number threshold, finding the best segmentation plane through a particle swarm optimization algorithm to divide the current node, and outputting a left and right node facet set as the left and right child nodes of the current node; continuing to recursively divide the divided child nodes as the current node until the number of facets in the current node is no greater than the preset number threshold, stopping the recursion, marking the current node as a leaf node and returning it; and outputting the hierarchical bounding box tree structure after the recursion ends.

[0077] Specifically, a BVH tree is constructed based on PSO.

[0078] like Figure 2a and Figure 2bAs shown, the present invention considers constructing a BVH tree in a scene containing multiple facets, so as to accelerate the process of ray tracing through the scene. The facets in the scene need to be stored in layers of recursive bounding boxes. Before calculating the intersection with the triangular surface of the object, it is first determined whether the ray intersects with its bounding box. If the bounding box has no intersection with the ray, then it is obvious that there will be no intersection with the triangular surface of the object. This avoids a large number of unnecessary intersection operations and greatly reduces the amount of calculation. The key lies in how to quickly construct a high-quality BVH tree.

[0079] First, the entire scene is input into the algorithm to parse the positions of all the faces in the scene; then a root node is created and all the faces are enclosed in a bounding box. Because if each leaf node contains only one face, the depth of the tree must be maximized, which not only takes a long time to build the tree, but also requires more visits to detect the face that finally intersects with the ray when traversing the tree structure, compared to the BVH tree whose leaf nodes contain multiple faces. Therefore, a reasonable number threshold needs to be set. , to limit the number of faces that can be contained in a leaf node. When the number of faces in a node is less than , stop the recursion immediately and return to the leaf node.

[0080] The flowchart of using PSO to construct the entire BVH tree is as follows Figure 3 shown.

[0081] The specific details of the algorithm for constructing the entire BVH tree using PSO are shown in Table 2:

[0082] Table 2

[0083] In one embodiment of the present invention, the ray tracing process is efficiently performed based on the constructed hierarchical bounding box tree structure. Specifically, the ray tracing engine will gradually traverse the three-dimensional scene along the pre-set ray direction, starting from the emission source. In this process, whenever the ray encounters a node of the BVH tree, it will first determine whether the ray intersects with the bounding box of the current node. If the ray does not intersect with the bounding box, the node and all its child nodes can be immediately excluded without further processing, thereby saving a lot of unnecessary computing resources.

[0084] If the ray intersects with the bounding box, the algorithm will recursively traverse the child nodes of the node. For each child node, the intersection check between the ray and the bounding box is also performed first. This process will continue until the ray intersects with the bounding box of a leaf node. At this time, the algorithm will perform accurate ray-geometry intersection checks on the specific geometry contained in the leaf node (such as the face of a building). If the ray intersects with a geometry, the physical effects of the ray such as reflection, refraction, and absorption are calculated based on the material, shape and other properties of the geometry, and the state of the ray (such as direction, energy, etc.) is updated.

[0085] In the entire ray tracing process, the use of the BVH tree structure can significantly reduce the number of intersection checks between rays and objects in the scene, thereby improving the efficiency of ray tracing. Especially in complex and changing scenes, such as urban buildings and mountainous terrain, the BVH tree structure can more effectively manage the geometry in the scene, ensuring that the ray tracing process is both fast and accurate, and providing high-quality data support for subsequent channel modeling, image rendering and other tasks.

[0086] Through the above method, simulation results are given to verify the effectiveness of the proposed method. The constructed ray tracing model and optimization algorithm are implemented based on Java. In the simulation, a simple urban scene is selected as the input of ray tracing. Figure 4a and Figure 4b As shown in the figure, the entire simulation scene contains 12 buildings, a total of 72 building surfaces, 144 edges, and the height of each building is between 10 and 50 (m). The area is approximately 200×300 The transmitting antenna is at (0.0, 0.0, 40.0), and the receiving antenna is at (80.0, 0.0, 0.0). The number of transmitted rays is 655362, which are evenly distributed in all directions in space.

[0087] In the scene of Figure 4, it takes 38935ms to directly traverse all rays and faces in the scene; it takes 11272ms to build a conventional BVH tree and perform ray tracing; it takes 9659ms to build a tree and perform ray tracing using the algorithm proposed in the present invention. In the scene of Figure 4, compared with the direct traversal ray tracing, the algorithm proposed in the present invention speeds up by 75.2%, and compared with the conventional BVH bounding box algorithm, the algorithm speeds up by 14.3%.

[0088] FIG5 shows a comparison of the running time of the algorithm proposed in the present invention and other comparison algorithms when the number of scene patches is fixed to 785 and the number of rays emitted at the transmitting antenna is changed. Figure 5aBy comparing the running time of the direct traversal ray tracing algorithm (Travel) with the other three acceleration algorithms, it can be seen that when the number of rays increases, the running time of the direct traversal algorithm increases linearly, which is much longer than the other three comparison algorithms. Figure 5b The running time of the three acceleration algorithms is compared. When the number of rays is small, the kdTree algorithm takes the shortest time, and its rising speed is also the fastest as the number of rays increases. Due to the advantage of the fast speed of the PSO tree building process, in the same scenario, the time consumption of this algorithm (SAH-PSO) is always less than the conventional BVH algorithm (BVH-SAH). As the number of rays reaches a large magnitude, the algorithm proposed in the present invention is superior to the other two acceleration algorithms.

[0089] FIG6 studies the time consumed by ray tracing using several algorithms when the number of emitted rays is fixed at 655362 and the scene scale, that is, the number of faces in the scene, is changed. Figure 6a A comparison was made between the direct traversal of the meshes for ray tracing (Travel) and the other three acceleration algorithms. When the number of meshes in the scene increases, its running time increases rapidly, far exceeding the other three acceleration algorithms. Figure 6b By comparing the running time of the PSO-SAH algorithm proposed in this invention with the kdTree algorithm and the SAH algorithm, it can be seen that the running time of the kdTree algorithm is longer than that of the other two acceleration algorithms. The running efficiency of this algorithm is better than that of the other two algorithms, and as the number of scene patches increases, its advantage of short tree building time is gradually reflected.

[0090] Figure 7 In the paper, by changing the number of facets in the scene, the time taken by the PSO-SAH algorithm and the SAH algorithm proposed in the present invention in a single search for the best splitting plane is compared. It can be seen that when the number of facets in the scene is small, the time taken by the two algorithms is almost the same. This is because the best splitting plane only exists in a very limited search space. Even if the best splitting plane is traversed, the time required is basically similar to that of PSO. As the number of facets in the scene increases, the search space gradually becomes larger, and traversal optimization becomes difficult. The huge amount of calculation leads to a sharp increase in running time. At this time, the advantages of this algorithm are gradually reflected, and the best splitting plane can be found in a shorter time.

[0091] In view of the problem that the BVH tree building speed based on the SAH algorithm is slow and it is difficult to perform ray tracing to establish a channel model in a short time, the present invention proposes an algorithm for SAH optimization based on a particle swarm algorithm to assist in establishing a BVH tree. Specifically, for emergency scenarios, channel modeling of the scene is required, and the ray tracing traversal algorithm takes too long, so an acceleration algorithm needs to be introduced. Although the conventional BVH bounding box algorithm can have a certain acceleration effect in the traversal process, its tree building process requires traversing the segmentation points to find the best segmentation plane, and the process is very time-consuming. Therefore, the present invention proposes an algorithm for finding the best segmentation plane using PSO. The simulation results show that the proposed algorithm has a good acceleration effect regardless of the increase in the number of scene patches or the number of rays, and has a certain performance improvement compared to its benchmark algorithm and other acceleration algorithms.

[0092] The fast deterministic channel modeling device based on improved bounding box ray tracing provided by the present invention is described below. The fast deterministic channel modeling device based on improved bounding box ray tracing described below and the fast deterministic channel modeling method based on improved bounding box ray tracing described above can be referred to each other. Figure 8 As shown, it includes the following modules: The surface area heuristic value acquisition module 810 is used to acquire a scene containing multiple buildings and determine the surface area heuristic value of a potential segmentation point in the scene; a potential segmentation point is a point used to divide the space represented by the current node into two subspaces during the process of constructing the hierarchical bounding box tree; a node is a unit in the hierarchical bounding box tree structure; An optimization problem modeling module 820 is used to obtain a set of position surface area heuristic values ​​of all potential segmentation points, so as to find a minimum surface area heuristic value in the set as an optimization problem for modeling; The hierarchical bounding box tree construction module 830 is used to obtain a preset threshold value. When the number of facets in the space represented by the node is greater than the threshold value, the optimization problem is solved based on the particle swarm optimization traversal segmentation point optimization method, the segmentation plane is determined, and the hierarchical bounding box tree structure is constructed.

[0093] Fig. 9 An example of a physical structure diagram of an electronic device is shown in FIG. Fig. 9As shown, the electronic device may include: a processor 910, a communication interface 920, a memory 930 and a communication bus 940, wherein the processor 910, the communication interface 920 and the memory 930 communicate with each other through the communication bus 940. The processor 910 may call the logic instructions in the memory 930 to execute a fast deterministic channel modeling method based on improved bounding box ray tracing, the method comprising: obtaining a scene containing multiple buildings, determining the surface area heuristic value of a potential segmentation point in the scene; a potential segmentation point is a point used to divide the space represented by the current node into two subspaces in the process of constructing a hierarchical bounding box tree; a node is a unit in the hierarchical bounding box tree structure; obtaining a set of position surface area heuristic values ​​of all potential segmentation points, so as to find the minimum surface area heuristic value in the set as an optimization problem for modeling; obtaining a preset threshold value, when the number of patches in the space represented by the node is greater than the threshold value, solving the optimization problem based on the particle swarm optimization traversal segmentation point optimization method, determining the segmentation plane, and constructing a hierarchical bounding box tree structure.

[0094] In addition, the logic instructions in the above-mentioned memory 930 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0095] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the fast deterministic channel modeling method based on improved bounding box ray tracing provided by the above methods, the method including: obtaining a scene containing multiple buildings, and determining the surface area heuristic value of a potential segmentation point in the scene; a potential segmentation point is a point used to divide the space represented by a current node into two subspaces during the process of constructing a hierarchical bounding box tree; a node is a unit in the hierarchical bounding box tree structure; obtaining a set of position surface area heuristic values ​​of all potential segmentation points to find the minimum surface area heuristic value in the set as an optimization problem for modeling; obtaining a preset threshold value, and when the number of patches in the space represented by the node is greater than the threshold value, solving the optimization problem based on a particle swarm optimization traversal segmentation point optimization method, determining the segmentation plane, and constructing a hierarchical bounding box tree structure.

[0096] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the fast deterministic channel modeling method based on improved bounding box ray tracing provided by the above-mentioned methods, the method comprising: obtaining a scene containing multiple buildings, and determining a surface area heuristic value of a potential segmentation point in the scene; a potential segmentation point is a point used to divide the space represented by a current node into two subspaces during the process of constructing a hierarchical bounding box tree; a node is a unit in the hierarchical bounding box tree structure; obtaining a set of position surface area heuristic values ​​of all potential segmentation points, so as to find the minimum surface area heuristic value in the set as an optimization problem for modeling; obtaining a preset threshold value, and when the number of patches in the space represented by the node is greater than the threshold value, solving the optimization problem based on a traversal segmentation point optimization method of particle swarm optimization, determining the segmentation plane, and constructing a hierarchical bounding box tree structure.

[0097] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative effort.

[0098] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fast deterministic channel modeling method based on improved bounding box ray tracing, characterized in that: include: Obtaining a scene containing multiple buildings, determining a surface area heuristic value of a potential segmentation point in the scene; The potential segmentation point is a point used to divide the space represented by the current node into two subspaces in the process of constructing the hierarchical bounding box tree; the node is a unit in the hierarchical bounding box tree structure; Obtaining a set of position surface area heuristic values ​​of all potential segmentation points, and modeling the method of finding a minimum surface area heuristic value in the set as an optimization problem; A preset threshold value is obtained. When the number of facets in the space represented by the node is greater than the threshold value, the optimization problem is solved based on a traversal segmentation point optimization method of particle swarm optimization, the segmentation plane is determined, and a hierarchical bounding box tree structure is constructed.

2. The fast deterministic channel modeling method based on improved bounding box ray tracing according to claim 1, characterized in that: The step of obtaining a scene containing multiple buildings and determining a surface area heuristic value of a potential segmentation point in the scene specifically includes: Obtain a scene containing multiple buildings, traverse each face in the scene and determine its position relative to a potential segmentation point; For each potential segmentation point, the time complexity of constructing a hierarchical bounding box tree using a surface area heuristic algorithm is determined according to the number of facets included in the potential segmentation point, and the time complexity is used as the surface area heuristic value of the segmentation point.

3. The fast deterministic channel modeling method based on improved bounding box ray tracing according to claim 1, characterized in that: The particle swarm optimization-based traversal segmentation point optimization method solves the optimization problem, specifically including: Get the current scene patch information and the bounding box volume list containing the patches as input for the particle swarm optimization algorithm; Determine the particle population size and number of iterations based on the scene size and number of patches; According to the input, the particle population size and the number of iterations, the potential segmentation points are traversed based on a particle swarm optimization algorithm to solve and determine an optimal solution.

4. The fast deterministic channel modeling method based on improved bounding box ray tracing according to claim 3 is characterized in that: The obtaining of the current scene patch information and the bounding box volume list including the patches specifically includes: Sort the n faces in the current scene in order according to their spatial positions and along a certain coordinate axis to obtain an ordered queue of all faces; Based on the ordered queue, a defined patch information list is obtained, and an information set of all patches is determined according to the patch information list; Based on the ordered queue, a left bounding box volume including the first i facets from left to right is defined, and a left sub-bounding box volume set is determined based on the left bounding box volume; a right bounding box volume including the first i facets from right to left is defined, and a right sub-bounding box volume set is determined based on the right bounding box volume.

5. The fast deterministic channel modeling method based on improved bounding box ray tracing according to claim 3, characterized in that: The step of traversing the potential segmentation points based on the particle swarm optimization algorithm according to the input, the particle population size and the number of iterations to solve and determine the optimal solution specifically includes: Initialize the population: Initialize a population of k particles, and randomly initialize the particle positions in the area where the surface area heuristic value is greater than the preset threshold; evaluate all particles, set the individual optimal position of each particle and the optimal position of the population; Evaluate particles: According to the position of each particle, obtain the index of the bounding box volume list corresponding to it, and calculate the fitness value of each particle based on the fitness function according to the left and right bounding box volumes corresponding to the index and the number of faces inside the bounding box; Update particle status: For each particle, if the current position fitness value is better than its individual optimal fitness, then update its individual optimal position; if the current position fitness value is better than its population optimal fitness, then update its population optimal position; The steps of evaluating particles and updating particle states are repeated until the optimal splitting plane is determined after the iterative convergence condition is met.

6. The fast deterministic channel modeling method based on improved bounding box ray tracing according to claim 1, characterized in that: The constructing of the hierarchical bounding box tree structure specifically includes: Analyze the positions of all faces in the scene; Create a current node and enclose all the faces of the current node with a bounding box; Determine the number of facets in the current node. When the number of facets in the current node is greater than a preset number threshold, find the best splitting plane to split the current node by using a particle swarm optimization algorithm, and output a left and right node facet set as the left and right child nodes of the current node. Continue to recursively divide the divided child node as the current node until the number of facets in the current node is no more than the preset number threshold, then stop the recursion and mark the current node as a leaf node and return; After the recursion is completed, the hierarchical bounding box tree structure is output.

7. The fast deterministic channel modeling method based on improved bounding box ray tracing according to claim 1, characterized in that: After modeling the optimization problem of finding the minimum surface area heuristic value in the set, the method further comprises: A preset threshold value is obtained, and when the number of facets in the space represented by the node is not greater than the threshold value, a surface area heuristic traversal method is used to find the best segmentation plane.

8. A fast deterministic channel modeling device based on improved bounding box ray tracing, characterized in that: include: A surface area heuristic value acquisition module, used for acquiring a scene containing multiple buildings and determining a surface area heuristic value of a potential segmentation point in the scene; The potential segmentation point is a point used to divide the space represented by the current node into two subspaces in the process of constructing the hierarchical bounding box tree; the node is a unit in the hierarchical bounding box tree structure; An optimization problem modeling module is used to obtain a set of position surface area heuristic values ​​of all potential segmentation points, and find a minimum surface area heuristic value in the set as an optimization problem for modeling; The hierarchical bounding box tree construction module is used to obtain a preset threshold value. When the number of facets in the space represented by the node is greater than the threshold value, the optimization problem is solved based on the particle swarm optimization traversal segmentation point optimization method, the segmentation plane is determined, and the hierarchical bounding box tree structure is constructed.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the fast deterministic channel modeling method based on improved bounding box ray tracing as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the fast deterministic channel modeling method based on improved bounding box ray tracing as claimed in any one of claims 1 to 7 is implemented.

Images