Method and system for adaptive adjustment of simulation accuracy based on multi-dimensional parameter linkage

Abstract

The application provides a visibility simulation precision self-adaptive adjustment method and system based on multi-dimensional parameter linkage, breaks through the conventional design idea of traditional visibility simulation fixed precision, single-dimensional parameter adjustment and only relying on instantaneous hardware state, constructs a two-way nonlinear self-adaptive precision regulation closed loop with global overall planning and local self-locking, abandons the one-size-fits-all static parameter allocation mode, dynamically configures the sampling step and grid LOD resolution according to the differences of different computing power working conditions and single line personalized characteristics, and considers the running stability, operation efficiency and complex scene shielding judgment precision of large-scale parallel simulation without additional hardware modification, and overcomes the inherent defects of the prior art that precision and efficiency cannot be considered, scene adaptability is poor, large-scale simulation is prone to freezing, and key areas are prone to missed judgment and misjudgment.

Description

Technical Field

[0001] This application relates to the field of simulation technology, and in particular to a method and system for adaptive adjustment of visual simulation accuracy based on multi-dimensional parameter linkage. Background Technology

[0002] Large-scale spatial visibility simulation technology is widely used in security monitoring, airspace line-of-sight detection, terrain visibility analysis, and scene visualization simulation. Its core principle is to rely on three-dimensional geographic data such as terrain models and building structures, and to complete the visibility simulation between observation points and target points through point-by-point elevation sampling and spatial occlusion determination. With the development of large-scale simulation scenarios and dense node deployment, existing visibility simulation adjustment schemes have gradually exposed many technical defects and are difficult to adapt to the engineering application requirements of large-scale parallel line-of-sight detection.

[0003] Current mainstream visual simulation methods mostly adopt a "one-size-fits-all" static simulation strategy with fixed resolution modeling and fixed sampling step size, which cannot dynamically adapt simulation accuracy according to real-time computing load and line-of-sight path characteristics. Some traditional solutions rely solely on line-of-sight distance for simple accuracy adjustment, resulting in a single adjustment dimension and extremely poor adaptability. They fail to take into account core influencing factors such as the computing congestion characteristics of large-scale simulations across the entire domain, the differentiated computing costs of each line of sight, the complexity of the terrain, and the priority of business security. This easily leads to problems such as computing power accumulation, screen stuttering, and computational latency jitter during large-scale parallel simulations, severely reducing simulation stability.

[0004] Meanwhile, existing technologies have significant shortcomings in terrain feature quantification and parameter fusion. They often use simple mean or extreme value methods to characterize terrain complexity, which fails to reflect the occlusion mechanism of "high sensitivity to near terrain and low impact from distant terrain" in real-world visibility scenarios. This can easily lead to the loss of occlusion features in complex micro-terrains and key security areas, resulting in missed or false occlusions. Furthermore, traditional parameter fusion methods often rely on simple linear weighted superposition, leading to severe homogenization of technical solutions and a lack of adaptive constraints. This makes it difficult to achieve a dynamic balance between computational efficiency and simulation accuracy, and to simultaneously meet the operational stability requirements of large-scale simulations and the need for high-precision occlusion determination.

[0005] Furthermore, conventional hardware adaptive adjustment methods often employ a single threshold and a coarse degradation strategy, which is susceptible to fluctuations in hardware parameters, resulting in accuracy oscillations and switching. They also fail to differentiate between bottleneck conditions related to memory, computing power, and bandwidth, leading to blind accuracy degradation, effective loss of shape features, and low hardware resource utilization. These shortcomings severely limit the practical application of large-scale, high-density visual simulation scenarios. Based on these deficiencies, there is an urgent need for a multi-dimensional parameter linkage method with strong adaptive capabilities that balances efficiency and accuracy in visual simulation accuracy adaptive adjustment. Summary of the Invention

[0006] Therefore, it is necessary to provide a method for adaptive adjustment of visual simulation accuracy based on multi-dimensional parameter linkage to address at least one of the above-mentioned technical problems, including: S1: The high-precision terrain model of the entire region is pre-divided into several standard geographic grid tiles, and multiple levels of LOD resolution are configured for each grid tile; S2: For the current simulation scenario, determine the global parameters used to characterize the overall simulation computing power pressure; S3: For each line of sight to be detected from the observation point to the target point, analyze and extract the set of all standard geographic grid tiles that the line of sight traverses and covers, and determine the local parameters used to characterize the unique computational overhead and environmental characteristics of each line of sight. S4: Integrates global and local parameters to independently configure a dedicated sampling step size and corresponding grid LOD resolution for each line of sight; S5: Based on the sampling step size and grid LOD resolution corresponding to each line of sight, load the grid tiles of the corresponding resolution and extract the elevation data point by point, complete the occlusion judgment in parallel, and output the visibility results of the observation point and the target point.

[0007] Further, step S2 includes: S21: Obtain the global simulation running status parameters, including the total number of global monitoring nodes, the total number of pending line requests in the current batch, the proportion of GPU parallel thread quota currently occupied by the system, and determine the global business coefficient that characterizes the instantaneous computing power pressure of the current simulation scenario; S22: Obtain all historical simulation thermal statistics parameters, including any one or more of the high-frequency calculation frequency of each region within the historical simulation period, historical batch line-of-sight congestion records, and historical GPU thread peak usage data, and determine the historical congestion correction factor. S23: Based on the historical congestion correction factor, dynamically correct the global business coefficients to determine the global parameters used for global accuracy constraints.

[0008] Furthermore, the global business coefficient is determined using formula (1): (1); The global parameters are determined using formula (2): (2); in, For global business coefficients, This represents the total number of monitoring nodes across the entire region. This represents the total number of lines of sight to be processed in the current batch. These are the system's preset number of reference nodes and reference line-of-sight, respectively. This represents the current percentage of GPU parallel thread quota usage. To preset the real-time parameter balancing weight coefficient, These are global parameters. To preset the historical thermal correction weighting coefficient, This is a historical congestion correction factor.

[0009] Further, step S3, the step of determining local parameters used to characterize the unique computational overhead and environmental characteristics of each line of sight, includes: S31: Obtain the security priority of all grid tiles in each line-of-sight grid tile set, and construct the security ladder constraint function; S32: Obtain the terrain undulation entropy of each tile in each view grid tile set to determine the comprehensive terrain undulation entropy of each view, and then determine the terrain computing power weight; S33: Determine the basic computing power of each line of sight based on the spatial straight-line distance between the observation point and the target point, the total number of grid tiles traversed by the line of sight, and the height difference of each line of sight; S34: Using the security weight ladder constraint function as a pre-constraint and the terrain computing power weight as a dynamic correction factor, the basic computing power of each line of sight is matched, and the local parameters of each line of sight are determined by coupling.

[0010] Furthermore, in step S32, the step of determining the integrated topographic relief entropy of each line of sight adopts formula (4): (4); in, The overall terrain undulation entropy for each line of sight, This represents the number of tiles in each set of tiles within the visual grid. For the first The topographic undulation entropy of a single tile For the first The cumulative distance of each tile from the observation point The total length of all lines of sight. This represents the average entropy of the tiles for each line of sight.

[0011] Furthermore, the local parameters are expressed using formula (5): (5); in, For local parameters, For the basic computing power of each line of sight, For terrain computing power weight, This is the security ladder constraint function.

[0012] Further, step S4 includes: S41: Pre-construct a multi-level global computing power threshold set [F1,F2], satisfying F1<F2; S42: When the global parameter ≥ F2, it is determined that the system is in a state of global computing power congestion, and the global priority balance strategy is triggered. With the global parameter as the core master control constraint variable, the upper limit of precision adjustment is constrained within the safe range of the lower limit of self-locking of each line-of-sight precision, the sampling step size is adaptively enlarged, and the mesh LOD resolution is lowered; S43: When the global parameter ≤ F1, it is determined that the system is in a state of idle computing power, and the global computing power forced balance constraint is lifted. With the local parameters of each line of sight as the precision control variables, the sampling step size and the mesh LOD resolution are differentially configured according to the numerical values of the local parameters; S44: When F1 < global parameter < F2, it is determined that the system is in a state of computing power transition. According to the global parameter and the local parameters, the sampling step size and the corresponding mesh LOD resolution of each line of sight are jointly determined.

[0013] Further, in step S44, the step of jointly determining the sampling step size of each line of sight adopts formula (6), which is expressed as: (6); Where, <http: / / www.w3.org / 1998 / Math / MathML>si< / http: / / www.w3.org / 1998 / Math / MathML>is the obtained sampling step size, <http: / / www.w3.org / 1998 / Math / MathML>s0< / http: / / www.w3.org / 1998 / Math / MathML>is the preset basic sampling step size of the system, <http: / / www.w3.org / 1998 / Math / MathML>g< / http: / / www.w3.org / 1998 / Math / MathML>is the global parameter, <http: / / www.w3.org / 1998 / Math / MathML>l< / http: / / www.w3.org / 1998 / Math / MathML>is the local parameter, <http: / / www.w3.org / 1998 / Math / MathML>kg< / http: / / www.w3.org / 1998 / Math / MathML>is the global adjustment coefficient, <http: / / www.w3.org / 1998 / Math / MathML>kl< / http: / / www.w3.org / 1998 / Math / MathML>is the local adjustment coefficient; The corresponding mesh LOD resolution is linked and matched with the sampling step size. The smaller the step size, the higher-order fine LOD mesh is matched, and the larger the step size, the lightweight low-order LOD mesh is matched.

[0014] Further, step S1 is specifically: uniformly grid-dividing the global high-precision terrain model in advance to generate standard geographic grid tiles of a unified specification; for each standard geographic grid tile, generating and storing multiple levels of decreasing LOD resolution levels in advance; Step S5 is specifically: under the GPU parallel computing architecture, according to the sampling step size and the mesh LOD resolution matched by each line of sight, complete the loading of the corresponding resolution grid tiles and the uniform elevation sampling operation, fuse the terrain elevation data and the building three-dimensional space data to carry out global space occlusion verification, and judge the occlusion of the line-of-sight link between the observation point and the target point one by one, and finally output the visible or invisible visibility simulation results corresponding to each line of sight On the other hand, the present application also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of any of the above methods are implemented.

[0015] This application provides a method and system for adaptive adjustment of visual simulation accuracy based on multi-dimensional parameter linkage. It breaks through the conventional design concept of fixed accuracy, single-dimensional parameter adjustment, and reliance on instantaneous hardware state in traditional visual simulation. It constructs a closed loop of bidirectional nonlinear adaptive accuracy control with global overall planning and local self-locking. It abandons the one-size-fits-all static parameter configuration mode and dynamically configures the sampling step size and grid LOD resolution according to different computing power conditions and individual characteristics of single line of sight. Without the need for additional hardware modification, it takes into account the running stability, computing efficiency and occlusion judgment accuracy of large-scale parallel simulation. It overcomes the inherent defects of existing technologies, such as the inability to balance accuracy and efficiency, poor scene adaptability, easy stuttering in large-scale simulation, and easy omission and misjudgment in key areas. Attached Figure Description

[0016] Figure 1 This is an application environment diagram of the visibility simulation accuracy adaptive adjustment method in one embodiment; Figure 2 This is a flowchart illustrating a method for adaptive adjustment of visibility simulation accuracy in one embodiment; Figure 3 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] This application provides a method for adaptive adjustment of visual simulation accuracy based on multi-dimensional parameter linkage, which can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. Server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0019] In one embodiment, such as Figure 2 As shown, a method for adaptive adjustment of visualization simulation accuracy based on multi-dimensional parameter linkage is provided, which can be used in scenarios such as simulation monitoring and search and rescue simulation. Specifically, it includes: S1: The high-precision terrain model of the entire region is pre-divided into several standard geographic grid tiles, and multiple levels of LOD resolution are configured for each grid tile; Specifically, the high-precision terrain model can be pre-divided into uniform grids to generate standard geographic grid tiles of uniform specifications. For each standard geographic grid tile, multiple levels of decreasing LOD resolution are pre-generated and stored. These levels can include the original high-precision resolution, 1 / 2 downsampling resolution, 1 / 4 downsampling resolution, etc., and the specific resolution can be arbitrarily set according to the accuracy and timeliness requirements of the current scene. Preferably, basic information such as terrain undulation features and business security level can also be pre-labeled on each grid tile for subsequent adaptive accuracy matching.

[0020] S2: For the current simulation scenario, determine the global parameters used to characterize the overall simulation computing power pressure; Specifically, global parameters representing the computational pressure of the simulation can be determined based on global information such as the total number of nodes and the total number of lines of sight in the current simulation scenario.

[0021] In a preferred embodiment, for the current large-scale parallel visual simulation scenario, the traditional single detection method that only collects instantaneous hardware load parameters can be abandoned. Instead, the real-time global simulation operation status and historical simulation thermal congestion patterns are combined to collect relevant parameters of global computing power operation in layers. Through a two-layer linkage mechanism of initial calculation of instantaneous computing power status and dynamic correction of historical congestion characteristics, global parameters that are used to accurately characterize the overall simulation comprehensive computing power pressure and adapt to the unified constraint of global accuracy can be obtained.

[0022] Specifically, step S2 may optionally include: S21: Obtain the global simulation running status parameters, including the total number of global monitoring nodes, the total number of pending line requests in the current batch, the proportion of GPU parallel thread quota currently occupied by the system, and determine the global business coefficient that characterizes the instantaneous computing power pressure of the current simulation scenario; Specifically, the global business coefficient can be determined using formula (1): (1); in, For global business coefficients, This represents the total number of monitoring nodes across the entire region. This represents the total number of lines of sight to be processed in the current batch. These are the system's preset number of reference nodes and reference line-of-sight, respectively. This represents the current percentage of GPU parallel thread quota usage. The preset real-time parameter balancing weight coefficient.

[0023] S22: Obtain all historical simulation thermal statistics parameters, including any one or more of the high-frequency calculation frequency of each region within the historical simulation period, historical batch line-of-sight congestion records, and historical GPU thread peak usage data, and determine the historical congestion correction factor. Specifically, the system can optionally calculate the cumulative line-of-sight calculation frequency for each geographic grid region within a preset historical simulation period, representing the region's inherent high-load, high-access computing power inertia; calculate the frequency and duration of congestion events such as the accumulation of multiple batches of concurrent line-of-sight requests and exceeding computational latency limits within the historical period, representing the congestion risk level of sudden batch simulations; and calculate the average peak sequence of GPU thread quota usage within the historical period, representing the long-term load capacity limit and overload potential of the hardware computing power. After normalizing the three types of heterogeneous parameters, a weighted average is used to determine the historical congestion correction factor.

[0024] S23: Based on the historical congestion correction factor, dynamically correct the global business coefficients to determine the global parameters used for global accuracy constraints.

[0025] Specifically, the global parameters can be determined using formula (2): (2); in, These are global parameters. For global business coefficients, To preset the historical thermal correction weighting coefficient, This is a historical congestion correction factor.

[0026] This embodiment provides a preferred implementation of step S2, which employs a two-layer global control logic of "real-time instantaneous computing power modeling + historical thermal congestion correction" to eliminate the random deviation of instantaneous state data, predict the overall simulation computing power congestion level, form a stable and accurate global precision constraint benchmark, and finally obtain complete global parameters for global precision constraints. Specifically, based on the three types of real-time state parameters in S21, global business coefficients are determined; then, based on the historical thermal statistics parameters in S22, historical congestion correction factors are introduced to dynamically correct the global business coefficients, resulting in global parameters for global precision constraints. This two-layer architecture can simultaneously consider the current instantaneous computing power load state of the simulation system and the historical congestion patterns of long-term operation, completely avoiding the technical defects of traditional single instantaneous load detection, such as biased judgment, delayed prediction, and high data randomness. Among them, the larger the value of the global parameter, the higher the comprehensive operating pressure of the current simulation scenario, specifically manifested in the large scale of the full-domain monitoring deployment, the large number of concurrent lines of sight to be detected, the saturation of GPU parallel thread resources, and the potential computing power risks of repeated calculations in high-frequency areas of the simulation and historical peak congestion. This application uses the computational congestion coefficient obtained from hierarchical iterative calculation as a unified global accuracy calibration benchmark, establishing a dynamic linkage mechanism between computational pressure and simulation accuracy. When the global parameter value is higher, i.e., the overall simulation computational load is closer to the threshold, the subsequent global accuracy adjustment benchmark can be appropriately relaxed, the basic accuracy threshold can be reasonably amplified, the independent computational overhead of a single line of sight can be distributed, and the computational power allocation ratio of global parallel simulation tasks can be balanced. Through this global adaptive adjustment mechanism, problems such as computational congestion, simulation screen stuttering, data computation latency jitter, and task backlog and crashes that are prone to occur in large-scale parallel line-of-sight detection are avoided from the root. It effectively balances computational efficiency and simulation accuracy in large-scale simulation scenarios, ensures the continuous, stable, and efficient steady-state operation of a large number of visual simulation tasks, and significantly improves the adaptability and robustness of the overall simulation system.

[0027] S3: For each line of sight to be detected from the observation point to the target point, analyze and extract the set of all standard geographic grid tiles that the line of sight traverses and covers, and determine the local parameters used to characterize the unique computational overhead and environmental characteristics of each line of sight. Specifically, a linear traversal algorithm can be used to analyze the path of each line of sight from the observation point to the target point, accurately screen and count all standard geographic grid tiles traversed by the spatial path of the line of sight, construct a dedicated tile set for each line of sight path, and lock the data calculation boundary for each line of sight. More specifically, the Bresenham linear traversal algorithm can be used to sample and traverse the spatial connection between the observation point and the target point grid cell by cell, accurately capture all standard geographic grid tiles penetrated and covered by the line of sight, remove invalid tiles that are tangent to the line of sight or do not actually penetrate, and generate a set of grid tiles for each line of sight without redundancy or omission. This provides accurate data boundary support for the subsequent accurate extraction, hierarchical determination and quantitative calculation of various parameters, and avoids parameter calculation deviations caused by interference from invalid tile data.

[0028] Then, the computing power of the line of sight, the specific parameters and dynamics of each tile in the line-of-sight-specific tile set are extracted to determine the local parameters of the unique computing overhead (computing power) and environmental characteristics (tile parameters) for each line of sight.

[0029] In a preferred embodiment, step S3, which involves determining local parameters to characterize the unique computational overhead and environmental characteristics of each line of sight, may optionally include: S31: Obtain the security priority of all grid tiles in each line-of-sight grid tile set, and construct the security ladder constraint function; Specifically, the preset security priority of all tiles in each line-of-sight grid tile set can be traversed. The priority can be divided into three levels: high, medium, and low, according to the business scenario. The highest security priority of all tiles in the path is selected as the security benchmark for this line of sight. Based on the hierarchical security threshold, a security ladder constraint function is constructed, which can be expressed as in formula (3) as a segmented step constraint form: (3); in, For security ladder constraint functions, The highest security priority is given to all tiles within each line of sight. This method employs progressively increasing precision constraint thresholds. It avoids the drawback of averaging, which weakens the protection capabilities of key areas, by using a security ladder constraint function. While seemingly simple, this function ensures that higher security levels across the entire line of sight correspond to smaller constraint thresholds, forcibly locking the lower limit of line-of-sight simulation precision. This prevents arbitrary increases in sampling step size or reduction in tile resolution in key security areas due to computational optimization, fundamentally eliminating the risk of missed detections due to occlusion in key areas and achieving differentiated precision safety net constraints for lines of sight at different security levels.

[0030] S32: Obtain the terrain undulation entropy of each tile in each view grid tile set to determine the comprehensive terrain undulation entropy of each view, and then determine the terrain computing power weight; Specifically, the terrain undulation entropy of each tile in the grid tile set of the line of sight can be extracted to quantify the terrain undulation changes, micro-topographic complexity and probability of local occlusion of a single tile. Then, by calculating the mean and other methods, the comprehensive terrain undulation entropy of the line of sight is determined, and the terrain computing power weight is determined for subsequent step 34 to perform computing power matching based on terrain characteristics.

[0031] In a preferred embodiment, the step of determining the integrated topographic undulation entropy for each line of sight can be performed using formula (4): (4); in, The overall terrain undulation entropy for each line of sight, This represents the number of tiles in each set of tiles within the visual grid. For the first The topographic undulation entropy of a single tile For the first The cumulative distance of each tile from the observation point The total length of all lines of sight. This is the average entropy of each line-of-sight tile (the average topographic undulation entropy of all tiles in the set of tiles in this line-of-sight grid).

[0032] In this embodiment, the comprehensive terrain undulation entropy of each line of sight is determined by combining the line-of-sight distance with the terrain undulation entropy of a single tile and the terrain variance. This aligns with the real occlusion mechanism of line-of-sight detection, where near terrain is highly sensitive and far terrain has a low impact. The influence of fine near terrain is strengthened by using distance weights, while the interference of flat terrain at a distance is weakened. At the same time, the constraint weight of the drastically undulating terrain section is amplified by the variance term, accurately depicting the real terrain complexity of the entire line of sight.

[0033] Furthermore, the solution can be based on the obtained line-of-sight integrated terrain undulation entropy. Construct the terrain complexity computing power weight, which can be represented as follows: In the formula, To preset the terrain suppression baseline coefficient, The terrain computing power weight is dynamic; the more dramatic the overall terrain undulation and the more complex the near-field micro-terrain, the stronger the suppression and constraint ability of the terrain computing power weight, so as to realize the accuracy self-adaptation and self-locking of complex terrain, and fundamentally avoid the problems of loss of near-field fine terrain occlusion features and misjudgment and missed judgment caused by simply reducing the computing power burden.

[0034] S33: Determine the basic computing power of each line of sight based on the spatial straight-line distance between the observation point and the target point, the total number of grid tiles traversed by the line of sight, and the height difference of each line of sight; Specifically, the selectable parameters include the linear distance L between the observation point and the target point, and the total number of effective grid tiles traversed by the line of sight. The maximum terrain elevation difference H along the line of sight determines the basic computing power for each line of sight. For example, the greater the viewing distance, the more tiles traversed, and the greater the terrain elevation difference, the higher the basic computing power for each line of sight. The larger the value, the higher the computational cost for interpolation calculation, elevation sampling, and occlusion determination for this line of sight, quantifying the differentiated computational burden for each line of sight.

[0035] S34: Using the security weight ladder constraint function as a pre-constraint and the terrain computing power weight as a dynamic correction factor, the basic computing power of each line of sight is matched, and the local parameters of each line of sight are determined by coupling.

[0036] Specifically, the local parameters can be expressed using formula (5): (5); in, For local parameters, For the basic computing power of each line of sight, For terrain computing power weight, This is the security ladder constraint function.

[0037] This embodiment presents a method for determining local parameters. It abandons the traditional simple linear weighted fusion approach and employs a multi-level nonlinear coupling mechanism combining pre-set step constraints, dynamic terrain suppression, and basic computing power adaptation to obtain unique, personalized local parameters for each line of sight. Specifically, a security step constraint function serves as the top-level mandatory constraint, locking the lower limit of accuracy; dynamic terrain computing power weights serve as intermediate correction factors, adaptively relaxing the computing power for suppressing complex terrain; and the basic computing power of the line of sight serves as the bottom-level benchmark, adapting to the inherent computational overhead of each line of sight. These three elements are nested and mutually constrained, achieving highly discriminative personalized parameter modeling for each line of sight. This approach, on the one hand, employs an extreme value-first, tiered constraint mode to completely resolve industry pain points such as weakened risk in key areas and missed detection of minor occlusions in complex terrain caused by traditional mean fusion, achieving precision self-locking assurance in high-security scenarios and complex terrains. On the other hand, it integrates multiple differentiated features such as spatial distance, tile traversal volume, terrain elevation difference, terrain undulation entropy, and security level from multiple dimensions, allowing the local parameters of each line of sight to be accurately matched with its own computing power consumption and business risk level. Finally, through a nonlinear nested coupling mechanism, it balances the security, accuracy, and efficiency of visual simulation, providing highly discriminative and non-homogeneous underlying feature support for subsequent high-precision adaptive coupling and parameter tuning of global and local parameters, significantly improving the overall stability and simulation accuracy of large-scale parallel visual simulation.

[0038] S4: Integrates global and local parameters to independently configure a dedicated sampling step size and corresponding grid LOD resolution for each line of sight; Specifically, it is optional to comprehensively consider global parameters and local parameters to construct a bidirectional interlocking non-linear coupling mechanism, adaptively determine the sampling step size and grid LOD resolution exclusive to each line of sight, and abandon the conventional control methods of traditional fixed-precision segmentation, single forward parameter adjustment, and linear weighted fusion.

[0039] In a preferred embodiment, step S4 may optionally include: S41: Pre-construct a multi-level global computing power threshold set [F1, F2], satisfying F1 < F2; in this way, the corresponding simulation system is divided into three computing power operating states: idle, transition, and congestion; furthermore, based on the current global parameters and local parameters, determine the corresponding sampling step size and grid LOD resolution.

[0040] S42: When F ≥ F2, it is determined that the system is in a state of global computing power congestion, trigger the global priority balance strategy, use the global parameter as the core main control constraint variable, and constrain the upper limit of precision adjustment within the safe range of the self-locking lower limit of each line of sight precision, adaptively increase the sampling step size, and lower the grid LOD resolution; S43: When F ≤ F1, it is determined that the system is in a computing power idle state,解除 the global computing power forced balance constraint, use the local parameters of each line of sight as the precision control variable, and differentially configure the sampling step size and grid LOD resolution according to the numerical size of the local parameters; in this way, high-overhead and high-risk line of sight high-precision adaptation and low-overhead and low-risk line of sight lightweight adaptation are achieved.

[0041] S44: When F1 < F < F2, it is determined that the system is in a computing power transition state (conventional or heavy load state), and联动 determine the sampling step size of each line of sight and the corresponding grid LOD resolution according to the global parameters and local parameters. More preferably, F3 may also be set between F1 and F2, satisfying F1 < F3 < F2; and the computing power transition state is further divided into a conventional state (F1 < F < F3) and a heavy load state (F3 < F < F2).

[0042] Specifically, the step of联动 determining the sampling step size of each line of sight according to the global parameters and local parameters may optionally adopt formula (6), expressed as: (6); Where, is the solved sampling step size, is the system preset basic sampling step size, is the global parameter, is the local parameter, is the global adjustment coefficient, is the local adjustment coefficient.

[0043] This formula forms a two-way balancing strategy: the global parameter positively controls the step size amplification, achieving heavy load reduction; the local parameter negatively constrains the step size, forcibly tightening the sampling accuracy for complex and high-risk lines of sight; the corresponding grid LOD resolution is linked and matched with the sampling step size, the smaller the step size matches the higher-order finer LOD grid, and the larger the step size matches the lightweight low-order LOD grid.

[0044] In this embodiment, a preferred embodiment of step S4 is given. In S41, by setting multi-level thresholds, the accuracy baseline of lines of sight at different risk levels can be differentially locked, thus preventing excessive degradation of accuracy and missed occlusion detection in high-value and highly complex scenarios. In the heavy-load game scenario of S42, the priority of global parameters is higher than the priority of personalized accuracy adjustment of local parameters of each line of sight. The accuracy adjustment range of all lines of sight is limited to above their respective accuracy self-locking lower limit. Simulation accuracy is uniformly and moderately reduced and the computational overhead of each line of sight is reduced only within a safe range. Priority is given to ensuring that large-scale batch simulation tasks do not experience congestion, lag, or crashes, achieving a steady-state balance of computing power across the entire domain. In the fine-adaptation scenario of S43, there is no computing power bottleneck in hardware resources, and the personalized adjustment capability of each line of sight is fully released. The larger the value of the local parameter, the higher the computational overhead of the line of sight, the higher the risk of terrain occlusion and the higher the level of security protection. The corresponding configuration is a smaller sampling step size and a higher-order fine LOD resolution. Conversely, the accuracy is moderately relaxed to achieve a personalized optimal match between simulation accuracy and computational efficiency. In the transitional dynamic adaptation scenario of S44, a two-parameter nonlinear coupling method of global + local is adopted to jointly determine the corresponding sampling step size and mesh LOD resolution. Specifically, the smaller the sampling step size, the higher the resolution level of the finer mesh tiles are matched; the larger the sampling step size, the lower the resolution level of the lightweight mesh tiles are matched.

[0045] This application employs a multi-level threshold grading architecture combined with global and local bidirectional interlocking nonlinear coupling for precision adaptive control, completely overcoming the technical shortcomings of traditional visual simulation, such as "fixed precision across the board, single-parameter linear tuning, and the contradiction between precision and efficiency." Through four-level hierarchical control of computing power conditions, differentiated control logic is achieved under different simulation loads. Heavy-load conditions rely on a global balancing mechanism to ensure stable operation of large-scale simulations; idle conditions rely on local features to achieve high-precision, refined detection; and transitional conditions achieve smooth dynamic adaptation through nonlinear linkage. Simultaneously, combined with a self-locking lower limit mechanism for the precision of each line of sight, it effectively avoids the problem of precision loss in key security areas and complex terrain areas caused by computing power optimization. While ensuring the efficiency and stability of large-scale parallel simulation operations, it maximizes the accuracy of occlusion detection in high-risk, high-value scenarios, achieving bidirectional optimal adaptation between global computing power balance and personalized precision for each line of sight. Compared to traditional single-parameter tuning schemes, it possesses extremely strong scene adaptability and technical superiority.

[0046] S5: Based on the sampling step size and grid LOD resolution corresponding to each line of sight, load the grid tiles of the corresponding resolution and extract the elevation data point by point, complete the occlusion judgment in parallel, and output the visibility results of the observation point and the target point.

[0047] Specifically, under a GPU parallel computing architecture, based on the sampling step size and grid LOD resolution matched for each line of sight, the corresponding resolution grid tile loading and uniform elevation sampling are completed. The terrain elevation data and building 3D spatial data are fused to carry out full-domain spatial occlusion verification. The occlusion judgment is performed on the line of sight between the observation point and the target point one by one, and finally the visibility simulation results corresponding to each line of sight are output.

[0048] More specifically, within a GPU parallel acceleration framework, for each line of sight to be detected, the adaptively configured sampling step size and grid LOD resolution from the aforementioned steps are retrieved. Only the target resolution grid tiles corresponding to the line of sight path are loaded into the video memory, avoiding the consumption of hardware resources by high-precision redundant tiles. Uniform sampling is performed along the spatial line of sight from the observation point to the target point according to the predetermined sampling step size, and terrain elevation data and building 3D contour data are extracted at each sampling point. Relying on the characteristics of parallel computing, spatial occlusion verification is carried out simultaneously on a large batch of line of sight sampling data. This not only determines natural landform occlusion caused by terrain undulations, but also matches the building 3D model to determine artificial building occlusion, completing the all-round occlusion detection of the entire line of sight. Finally, the visibility binary simulation results of each line of sight are output in batches, forming large-scale visibility simulation output data.

[0049] In this embodiment, step S5 relies on the GPU parallel architecture to achieve synchronous computation of large-scale line-of-sight occlusion judgments, fully matching the multi-dimensional adaptive precision configuration mechanism described above, and realizing a closed-loop operation mode of "differentiated precision parameter configuration and parallel unified detection". On the one hand, elevation sampling and occlusion judgment are carried out based on the adapted precise precision parameters, avoiding the loss of details or computational redundancy caused by fixed precision detection, effectively improving the accuracy and reliability of each line-of-sight occlusion judgment; on the other hand, through the parallel batch processing mechanism, the computational efficiency of large-scale, high-density visibility simulation scenarios is greatly improved, solving the shortcomings of traditional linear detection which is inefficient and cannot adapt to large-scale simulation business. At the same time, the dual occlusion judgment dimensions of terrain and buildings are integrated to cover the entire spatial occlusion scenario, avoiding the judgment omissions caused by single terrain detection, comprehensively improving the completeness and scene adaptability of visibility simulation results, and ensuring the stable, accurate, and efficient output of visibility results in large-scale simulation scenarios.

[0050] In summary, this application provides a method for adaptive adjustment of visibility simulation accuracy based on multi-dimensional parameter linkage. Through a complete closed-loop technical solution involving global terrain model tile hierarchical preprocessing, global computational pressure parameter acquisition, extraction of local feature parameters for each line of sight, fusion of global and local multi-dimensional parameters for differentiated parameter allocation, and accuracy-adaptive simulation calculations, this method effectively overcomes the inherent defects of existing visibility simulation technologies, such as fixed accuracy configuration, single adjustment dimension, and poor scene adaptability. The technical effects include at least the following: (1) This application abandons the traditional simulation scheme of uniformly fixing the sampling step size and model resolution in a static mode. It relies on multi-level LOD grid tiles to pre-build a basic adaptation carrier, providing a layered accuracy adaptation basis for simulation scenarios with different complexity and different computing power conditions. This effectively solves the problem that a single accuracy model cannot take into account the terrain details and computational efficiency of a large-scale simulation scenario, and greatly improves the scene adaptation breadth of visual simulation.

[0051] (2) This application innovatively distinguishes between the global computing power characteristics of the entire domain and the local differential characteristics of each line of sight. It combines the overall simulation computing power pressure with the exclusive computing overhead and environmental characteristics of each line of sight to achieve multi-dimensional parameter linkage control. This breaks the limitation of the traditional method of adjusting the accuracy by relying solely on the single distance parameter of the line of sight. It can accurately adapt to the differentiated computing needs of different lines of sight in large-scale parallel line-of-sight detection scenarios, avoid the problems of uneven computing power distribution and poor simulation stability caused by the concurrency of a large number of simulation tasks, and significantly improve the overall steady-state operation capability of large-scale visual simulation.

[0052] (3) This application independently configures a dedicated sampling step size and grid LOD resolution for each line of sight to be detected, realizing personalized adaptive adjustment of each line of sight and single precision. It can match the optimal simulation precision according to the real environmental characteristics and computational overhead of different lines of sight, avoiding the waste of computing power caused by the redundancy of precision in low-complexity scenes, and avoiding the occlusion judgment deviation caused by insufficient precision in high-complexity scenes, effectively balancing the computational efficiency and detection accuracy of visibility simulation.

[0053] (4) Based on the adapted precision parameters, this application completes grid tile loading, elevation data extraction and parallel occlusion judgment. Relying on the refined and differentiated precision adaptation mechanism, it ensures the reliability and accuracy of the final visibility results, adapts to large-scale visual simulation business scenarios with large range, high density and multiple nodes, and effectively makes up for the technical shortcomings of traditional static simulation solutions that are difficult to adapt to dynamic, large-scale and differentiated simulation needs.

[0054] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0055] In one embodiment, this application also provides a computer device, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores simulation execution data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a method for adaptive adjustment of visual simulation accuracy based on multi-dimensional parameter linkage.

[0056] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0057] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0058] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0059] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification. The above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application. It should be noted that for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for adaptive adjustment of visual simulation accuracy based on multi-dimensional parameter linkage, characterized in that, The method includes: S1: The high-precision terrain model of the entire region is pre-divided into several standard geographic grid tiles, and multiple levels of LOD resolution are configured for each grid tile; S2: For the current simulation scenario, determine the global parameters used to characterize the overall simulation computing power pressure; S3: For each line of sight to be detected from the observation point to the target point, analyze and extract the set of all standard geographic grid tiles that the line of sight traverses and covers, and determine the local parameters used to characterize the unique computational overhead and environmental characteristics of each line of sight. S4: Integrates global and local parameters to independently configure a dedicated sampling step size and corresponding grid LOD resolution for each line of sight; S5: Based on the sampling step size and grid LOD resolution corresponding to each line of sight, load the grid tiles of the corresponding resolution and extract the elevation data point by point, complete the occlusion judgment in parallel, and output the visibility results of the observation point and the target point.

2. The method according to claim 1, characterized in that, Step S2 includes: S21: Obtain the global simulation running status parameters, including the total number of global monitoring nodes, the total number of pending line requests in the current batch, the proportion of GPU parallel thread quota currently occupied by the system, and determine the global business coefficient that characterizes the instantaneous computing power pressure of the current simulation scenario; S22: Obtain all historical simulation thermal statistics parameters, including any one or more of the high-frequency calculation frequency of each region within the historical simulation period, historical batch line-of-sight congestion records, and historical GPU thread peak usage data, and determine the historical congestion correction factor. S23: Based on the historical congestion correction factor, dynamically correct the global business coefficients to determine the global parameters used for global accuracy constraints.

3. The method according to claim 2, characterized in that, The global business coefficient is determined using formula (1): （1）； The global parameters are determined using formula (2): （2）； in, For global business coefficients, This represents the total number of monitoring nodes across the entire region. This represents the total number of lines of sight to be processed in the current batch. These are the system's preset number of reference nodes and reference line-of-sight, respectively. This represents the current percentage of GPU parallel thread quota usage. To preset the real-time parameter balancing weight coefficient, These are global parameters. To preset the historical thermal correction weighting coefficient, This is a historical congestion correction factor.

4. The method according to claim 1, characterized in that, Step S3, which involves determining local parameters to characterize the unique computational overhead and environmental features of each line of sight, includes: S31: Obtain the security priority of all grid tiles in each line-of-sight grid tile set, and construct the security ladder constraint function; S32: Obtain the terrain undulation entropy of each tile in each view grid tile set to determine the comprehensive terrain undulation entropy of each view, and then determine the terrain computing power weight; S33: Determine the basic computing power of each line of sight based on the spatial straight-line distance between the observation point and the target point, the total number of grid tiles traversed by the line of sight, and the height difference of each line of sight; S34: Using the security weight ladder constraint function as a pre-constraint and the terrain computing power weight as a dynamic correction factor, the basic computing power of each line of sight is matched, and the local parameters of each line of sight are determined by coupling.

5. The method according to claim 4, characterized in that, In step S32, the step of determining the integrated topographic relief entropy of each line of sight is carried out using formula (4): （4）； in, The overall terrain undulation entropy for each line of sight, This represents the number of tiles in each set of tiles within the visual grid. For the first The topographic undulation entropy of a single tile. For the first The cumulative distance of each tile from the observation point The total length of all lines of sight. This represents the average entropy of the tiles for each line of sight.

6. The method according to claim 4, characterized in that, The local parameters are expressed by formula (5): （5）； in, For local parameters, For the basic computing power of each line of sight, For terrain computing power weight, This is the security ladder constraint function.

7. The method according to claim 1, characterized in that, Step S4 includes: S41: Pre-construct a multi-level global computing power threshold set [F1,F2], satisfying F1<F2; S42: When the global parameter is greater than or equal to F2, the system is determined to be in a state of global computing power congestion. The global priority balance strategy is triggered. The global parameter is the core main control constraint variable. Within the safe range of the self-locking lower limit of each line of sight accuracy, the upper limit of the constraint accuracy is adjusted. The sampling step size is adaptively increased and the grid LOD resolution is reduced. S43: When the global parameter ≤ F1, it is determined that the system is in a computing power idle state, the global computing power forced balance constraint is解除 (should be "解除", translated as "removed"), and each line-of-sight local parameter is used as the precision control variable. The sampling step size and grid LOD resolution are differentially configured according to the numerical size of the local parameter. S44: When F1 < global parameter < F2, it is determined that the system is in a computing power transition state. According to the global parameter and local parameter, the sampling step size and the corresponding grid LOD resolution of each line of sight are联动确定 (should be "联动确定", translated as "jointly determined") in a coordinated manner.

8. The method according to claim 7, characterized in that, In step S44, the step of jointly determining the sampling step size of each line of sight adopts formula (6), which is expressed as: （6）； in, To obtain the sampling step size, The system is preset with a basic sampling step size. These are global parameters. For local parameters, This is the global adjustment coefficient. This is the local adjustment coefficient; The corresponding grid LOD resolution is matched with the sampling step size in a coordinated manner. The smaller the step size, the higher-order fine LOD grid is matched, and the larger the step size, the lightweight low-order LOD grid is matched.

9. The method according to any one of claims 1 to 8, characterized in that, Step S1 is specifically as follows: First, the全域高精度地形模型 (should be "全域高精度地形模型", translated as "global high-precision terrain model") is evenly grid-divided to generate standard geographic grid tiles of a unified specification; for each standard geographic grid tile, a multi-level decreasing LOD resolution level is pre-generated and stored. Step S5 is specifically as follows: Under the GPU parallel computing architecture, according to the sampling step size and grid LOD resolution matched by each line of sight, the grid tile loading and uniform elevation sampling operations corresponding to the resolution are completed, the terrain elevation data and the three-dimensional spatial data of the building are融合 (should be "融合", translated as "integrated") to carry out global spatial occlusion verification, and the occlusion judgment of the line-of-sight link between the observation point and the target point is carried out one by one. Finally, the visible or invisible visibility simulation results corresponding to each line of sight are output.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it realizes the steps of the method described in any one of claims 1 to 9.

Images