Hierarchical loading and rendering method for three-dimensional geographic space data

Abstract

The invention relates to the technical field of virtual reality processing software, and discloses a hierarchical loading and rendering method for three-dimensional geographic space data, which comprises the following steps: extracting a displacement vector and a rotation deflection amount of a virtual camera, calculating a weight sensitive field angle according to the displacement vector and the rotation deflection amount, and constructing a rendering weight distribution area in a view cone, the method comprises the following steps: determining a rendering loading weight according to a deviation angle of a geographic entity node relative to a visual axis, distributing an asynchronous scheduling queue by utilizing the rendering loading weight, matching detail levels, monitoring time consumption of a rendering pipeline frame in real time, executing grading and degrading processing when the time consumption exceeds a threshold value, and calling a rendering shader to execute linear interpolation. According to the method, resource allocation is dynamically adjusted according to viewpoint motion features, the problem of data loading lag in high-dynamic interaction is solved, continuous details of a visual center area are guaranteed through a weight-sensitive closed-loop degradation mechanism, geometric jump is eliminated, and the rendering stability of a three-dimensional geographic scene is improved.

Description

Technical Field

[0001] This invention relates to a hierarchical loading and rendering method for three-dimensional geospatial data, belonging to the field of virtual reality processing software technology. Background Technology

[0002] With the popularization of real-scene 3D and digital twin technologies, real-time high-precision rendering of 3D geospatial data has become a key support for virtual reality processing software. The industry generally adopts a level-by-level loading method based on Euclidean distance to complete the asynchronous scheduling and on-demand loading of tile data according to the spatial distance between geographic entity nodes and virtual camera viewpoints. When users perform fast viewpoint translation or non-linear rotation, the distance-based spatial metric ignores the logical relationship between motion vectors and data loading rate. Geographic entity nodes that are about to move out of the field of view occupy bus bandwidth and video memory space, causing the bandwidth of the central area of ​​the field of view to be squeezed by redundant edge requests, resulting in model jumps or texture loading lags, and a decrease in the determinism of the graphics processing unit during dynamic interaction.

[0003] Industry attempts to increase prefetch depth or expand hardware cache capacity to compensate for latency and improve system power consumption cannot solve the displacement deviation between the perception center of gravity and the loading center of gravity. Existing rendering optimization solutions have limitations. For example, Chinese invention patent CN110796722B discloses a 3D rendering method and device. It improves rendering efficiency by counting the number of rendering pipeline nodes and establishing a resource sharing mechanism between nodes to coordinate the work within the pipeline. Essentially, it is a passive resource scheduling for static architectures or low-speed scenes. Its core defect is the lack of active perception and feedforward response capability for the high-frequency motion characteristics of virtual cameras. When faced with highly dynamic interactive scenarios such as users performing fast scanning or high-speed cruising, this strategy, which only focuses on the collaboration of nodes within the pipeline and ignores the evolution trend of viewpoint pose, cannot reconstruct the weight distribution in real time based on the viewpoint rotation and deflection. As a result, within millisecond-level frame slots, computing and bandwidth resources cannot be accurately focused on the core visual area, making it difficult to fundamentally eliminate visual flickering and stuttering caused by data loading timing misalignment.

[0004] Therefore, how to construct a perception weight evaluation mechanism based on motion vector compensation so that the dynamic delivery of rendering resources is consistent with the viewpoint interaction trend is the technical problem to be solved by this invention. Summary of the Invention

[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A hierarchical loading and rendering method for three-dimensional geospatial data, comprising the following steps: Step S101: Obtain the real-time pose matrix of the virtual camera, and extract the motion state parameters of the virtual camera based on the real-time pose matrix. The motion state parameters include displacement vector and rotation deflection. Step S102: Based on the numerical changes of displacement vector and rotation deflection, calculate the weight-sensitive angle of the virtual camera in the current rendering cycle, and construct a rendering weight distribution area with the view axis as the center axis within the view frustum of the virtual camera. Step S103: Obtain the spatial location data of the geographic entity node to be loaded, calculate the offset angle of the geographic entity node relative to the view axis, and determine the rendering loading weight of the geographic entity node based on the ratio of the offset angle to the weight-sensitive angle. Step S104: Based on the rendering loading weight, the geographic entity nodes are assigned to the corresponding asynchronous scheduling queue, and the corresponding level of detail is matched according to the size of the rendering loading weight to perform geospatial data parsing with the corresponding precision. Step S105: Monitor the frame latency of the rendering pipeline in real time. When the frame latency exceeds a preset latency threshold, perform graded degradation processing on the detail level of the geographic entity nodes in the asynchronous scheduling queue according to the rendering loading weight from small to large, until the output frame rate of the rendering pipeline reaches a preset stability threshold, wherein the stability threshold is not lower than... ; Step S106: Invoke the rendering shader of the graphics processor to perform geometric feature linear interpolation on the geographic entity nodes that are in the process of switching levels of detail, so as to smooth the geometric feature jumps in the process of switching levels of detail.

[0006] Preferably, determining the rendering loading weight of the geographic entity node in step S103 includes: comparing the offset angle with the weight-sensitive angle; if the offset angle is greater than the weight-sensitive angle, the geographic entity node is determined to be in a low-weight area, and the rendering loading weight is determined based on a preset minimum weight value; if the offset angle is less than or equal to the weight-sensitive angle, the rendering loading weight is calculated based on the ratio of the offset angle to the weight-sensitive angle using a preset weight decay function.

[0007] Preferably, step S101 involves extracting motion state parameters of the virtual camera, including: calculating the pose transformation matrix between the current rendering cycle and the previous rendering cycle, and decomposing the pose transformation matrix to obtain the displacement vector representing spatial translation and the rotation deflection amount representing the change in the view axis rotation angle.

[0008] Preferably, step S105 performs a graded degradation process, including: identifying a set of non-core nodes whose rendering loading weight is lower than a preset weight threshold; sending a scheduling suspension instruction through an asynchronous scheduling queue to stop asynchronous I / O requests for texture map data in the set of non-core nodes, and forcibly limiting the set of non-core nodes to only rendering a preset basic level of detail model.

[0009] Preferably, in step S104, allocating geographic entity nodes to corresponding asynchronous scheduling queues includes: dividing all geographic entity nodes within the view frustum into high-priority queues, medium-priority queues, and low-priority queues based on rendering loading weights; and allocating system bus bandwidth resources and video memory addressing space to each queue in descending order of priority.

[0010] Preferably, in step S106, geometric feature linear interpolation is performed, characterized in that the linear interpolation follows the following calculation rules: ,in, To obtain the geometric feature values ​​of geographic entity nodes after performing linear interpolation. These are the geometric feature values ​​of geographic entity nodes at a low level of detail. These are the geometric feature values ​​of geographic entity nodes at a high level of detail. This is a preset blending factor representing the timing progress of the switching of detail levels.

[0011] Preferably, the calculation of the weight-sensitive angle in step S102 includes: obtaining the real-time angular velocity of the virtual camera, and when the real-time angular velocity increases, synchronously decreasing the value of the weight-sensitive angle so that the rendering weight distribution area shrinks towards the central axis region.

[0012] Preferably, it further includes: introducing a scheduling de-jitter step based on historical weights, by storing the historical stable weights of geographic entity nodes within a preset number of preceding frames, and calculating the real-time change rate of rendering loading weights relative to historical stable weights; if the real-time change rate is lower than a preset gating threshold, then the current level of detail of the geographic entity node is maintained.

[0013] Preferably, the time delay threshold in step S105 is limited to: The frame latency is determined by monitoring the closed-loop latency of the rendering pipeline from task submission to frame buffer rendering.

[0014] Preferably, step S104 involves performing geospatial data parsing with the corresponding precision, including: retrieving the corresponding grid topology data and texture resources from the external storage medium according to the level of detail, and injecting the grid topology data and texture resources into the video memory buffer using a multi-threaded asynchronous I / O mechanism.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. In 3D geospatial data, motion tensors are extracted in real time through the pose matrix of virtual cameras, enabling the system to have scheduling capabilities that match the dynamic interaction rhythm when processing large-scale geospatial data. Real-time calculation of displacement vectors and rotation components allows the rendering pipeline to allocate bandwidth and video memory resources to geographic entity nodes in the core area of ​​the motion path, avoiding invalid input and output throughput in the areas to be moved out of the field of view, thereby eliminating graphic jumps caused by data loading lag during dynamic translation or rotation.

[0016] 2. By combining the perception weight score with the feedback suppression of the asynchronous scheduling queue, the system achieves stable output of the rendering pipeline load under complex interactive conditions. When the real-time bandwidth load approaches the critical value, the system performs non-linear hierarchical degradation based on the perception weight of the geographic entity node. By actively limiting the data parsing depth of low perception value areas, the system prioritizes the high-precision presentation of visually locked targets, ensuring that hardware resource utilization and user interaction needs remain dynamically consistent in the time domain. The system utilizes the graphics processor's shaders to perform linear interpolation on geometric feature differences within a preset time window, combined with a perceptual inertia-based scheduling de-jittering mechanism, to eliminate discrete jumps during hierarchical switching. The introduction of historically stable weights and the gating judgment of momentum thresholds avoid scheduling logic oscillations caused by sensor perturbations, ensuring that geographic entities exhibit visual geometric stability during the refinement evolution process, and solving the visual flickering caused by frequent switching of rendering states in high-frequency micro-motion environments.

[0017] 3. By coordinating angular acceleration compensation and texture sampling control, the rendering strategy adapts to the physiological characteristics of the human visual system. Under high-speed rotation, the system actively limits the maximum value of texture detail level, redirecting the saved graphics processing unit resources to the center of the field of view pointed to by the motion vector. It utilizes the persistence of vision to achieve reasonable truncation of computational precision, avoiding the ineffective computational waste of high-frequency texture scheduling under extreme dynamic conditions. It uses the depth buffer data of the previous rendering frame to extract the depth reference value of the spatial partition, and combines it with the motion tensor to achieve visibility prediction, reduce the prefetching priority of occluded areas, and cross-couple the depth probability map with the perception weight to avoid the invalid loading of invisible high-precision data, thereby improving the operational stability of the rendering pipeline in high-density scenes. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the hierarchical loading process for viewpoint motion features and frame latency feedback in this invention. Figure 2 This is a diagram showing the multi-module collaborative architecture and data interaction topology of the rendering system of this invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments. It should be noted that, in the absence of conflict, the various embodiments and technical features in the embodiments described below can be combined with each other. The following embodiments are intended to explain and illustrate the present invention, and are not intended to limit the scope of protection of the present invention.

[0020] A hierarchical loading and rendering method for 3D geospatial data dynamically adjusts rendering resource allocation based on the real-time motion characteristics of a virtual camera to address the data loading lag issue in virtual reality interaction. Applied to virtual reality processing software, it ensures the continuity of details in the visual center region by constructing a weight-sensitive closed-loop degradation mechanism. During the virtual camera motion state parameter extraction stage, the system acquires the virtual camera's pose matrix in real time and extracts its motion state parameters, including displacement vectors and rotational deflections. The system calculates the pose transformation matrix between the current rendering cycle and the previous rendering cycle and decomposes the pose transformation matrix to obtain displacement vectors representing spatial translation and rotational deflections representing changes in the view axis rotation angle. When the virtual camera... sampling period within edge Positive axis movement Unit and around Axis rotation At that time, the displacement vector is the translational increment, and the rotational deflection corresponds to the rotational angular displacement. To address the perception redundancy under high-speed motion, the system calculates the weighted sensitive angle of the virtual camera in the current rendering cycle based on the numerical changes of the displacement vector and rotational deflection. And construct a rendering weight distribution area centered on the view axis within the view frustum of the virtual camera; in determining At that time, the system obtains the real-time angular velocity of the virtual camera. ,when As it increases, it decreases simultaneously. The value of causes the rendering weight distribution area to shrink towards the central axis area. At this time, the high-definition rendering area is concentrated in the center of the field of view, and the visually blurred area at the edge of the field of view due to the rapid angular displacement is defined as the low weight area.

[0021] During the rendering and loading weight determination phase of geographic entity nodes, the system acquires the spatial location data of the geographic entity nodes to be loaded and calculates the offset angle of the geographic entity nodes relative to the view axis. The geographic entity node center projection mapping procedure is executed to transform the bounding box center point coordinates of the geographic entity node to the virtual camera's view space coordinate system. The offset angle is determined by calculating the dot product of the node direction vector and the view axis vector in the view space coordinate system. The numerical value; where the geographic entity node has a pre-defined data structure containing node identifier, spatial boundary radius, and multi-level detail index, the spatial boundary radius is used in calculating the offset angle. During the view frustum culling prediction, if the bounding sphere of a geographic entity node is completely outside the view frustum, the system will force the rendering to load weights. Set to zero; the system will and Compare: If Greater than The system determines that a geographic entity node is in a low-weight zone based on a preset minimum weight value; if... Less than or equal to Then according to and The proportional relationship is used to calculate the rendering loading weight through a preset weight decay function. The weight decay function is implemented through a discretized lookup table mapping logic. The quotient of the offset angle divided by the weight-sensitive subtended angle is used as an index to retrieve the coefficient from a preset zero-to-one-step incremental mapping table. When the quotient is zero, the weight is set to 0.1. As the quotient gradually increases, the weight decreases linearly at a rate of 0.1 weight levels per 0.1 unit increment of the quotient, until the quotient is 1, at which point the weight drops below 0.1. The weight decay function reaches its maximum value at the central axis and continues to decrease accordingly. The increase is not a linear decrease; based on the determined The system assigns geographic entity nodes to corresponding asynchronous scheduling queues and matches them with the corresponding level of detail to perform geospatial data parsing with the corresponding precision. The asynchronous scheduling queues are divided into high-priority queues, medium-priority queues, and low-priority queues. The system allocates system bus bandwidth resources and video memory addressing space to each queue in descending order of priority. When performing geospatial data parsing with the corresponding precision, the system retrieves the corresponding mesh topology data and texture resources from the external storage medium according to the level of detail. Using a multi-threaded asynchronous input / output mechanism, the mesh topology data and texture resources are injected into the video memory buffer.

[0022] To ensure deterministic rendering during dynamic interactions, the system monitors the frame latency of the rendering pipeline in real time. ,when Exceed When the preset delay threshold is met, according to In ascending order, the detail level of geographic entity nodes in the asynchronous scheduling queue is processed in a hierarchical degradation manner until the output frame rate of the rendering pipeline reaches at least [a certain value]. The system identifies the preset stability threshold; when performing graded degradation processing, the system recognizes... For non-core node sets below a preset weight threshold, a scheduling suspension command is sent via the asynchronous scheduling queue to abort asynchronous requests for texture map data in the non-core node set, and to force the non-core node set to render only the preset basic level of detail model; for geometric feature jumps during the level of detail switching process, the system calls the graphics processor's rendering shader to perform linear interpolation of geometric features on geographic entity nodes undergoing level switching; the linear interpolation follows the following calculation rules: ,in, To obtain the geometric feature values ​​of geographic entity nodes after performing linear interpolation. These are the geometric feature values ​​of geographic entity nodes at a low level of detail. These are the geometric feature values ​​of geographic entity nodes at a high level of detail. The preset blending factor represents the timing progress of the detail level switching, and its value range is: to In actual operation, the system introduces a scheduling de-jitter step based on historical weights; the system stores the historical stable weights of geographical entity nodes within a preset number of preceding frames. And calculate the current Compared to Real-time rate of change ,like If the threshold is lower than the preset threshold, the current level of detail for the geographic entity node is maintained; if... If the inertial region defined by the motion tensor is broken, then update... This triggers a change in the priority of tiered loading, thereby ensuring that the rendering state transitions are consistent with the user's intent.

[0023] Example 1: In a digital twin scene containing a high-density real-world 3D building complex, a virtual camera uses... The translational speed runs and is accompanied by Axial jitter; due to the amount of tile data scheduling requests triggered per unit time exceeding the physical bandwidth of the system bus, texture loading lag and model detail jumps occur in the center region of the field of view; to address the above conditions, the system extracts the pose matrix of the virtual camera and calculates the displacement vector and rotation deflection, and when the angular velocity is detected... achieve At that time, the weight-sensitive angle will be... from Adjust to and offset angle Greater than Rendering loading weight of geographic entity nodes The minimum weight value is determined; geographic entity nodes located in the center of the view axis are assigned to a high-priority queue, and their asynchronous read requests for mesh topology data and texture resources are processed first. By dynamically shrinking the weight distribution area and reconstructing the priority of the asynchronous scheduling queue, bandwidth resources are allocated to the central area of ​​perception.

[0024] When the virtual camera's real-time rate of change Falling back to Within the de-jitter gating threshold, and the frame latency of the rendering pipeline. Stable to At the same time, the level of detail of geographic entity nodes is smoothly restored through geometric feature linear interpolation, maintaining the output frame rate of the entire rendering pipeline at a certain level. The above stability thresholds ultimately eliminate visual flicker caused by the discrete degradation of detail levels, while ensuring the high fidelity of the viewpoint region model.

[0025] Example 2: In an experimental environment verifying the stability of large-scale 3D geographic scene rendering, the system runs on a virtual reality computing platform with graphics hardware acceleration capabilities, and its memory bandwidth limit is set to [value missing]. It also supports multi-threaded asynchronous addressing; the experiment used a resolution of Using a real-world 3D city dataset as the data source, this experiment simulates the pose evolution of a virtual camera under high dynamic interaction. The experiment monitors the texture reconstruction latency of geographic entity nodes in the central region of the field of view and the output frame rate of the rendering pipeline. During parameter calibration, the system determines the weighted sensitive angle. Mapping logic: Obtain the real-time angular velocity of the virtual camera. ,according to The numerical change dynamically searches for the aperture closure point; when In to When the interval is within, Determined as The maximum value; when Exceed After the threshold is reached, the system executes a linear contraction strategy, i.e. Follow As it increases, it decreases until it contracts to The lower limit value; to verify the stability of this decision logic in the engineering field, during the experiment, a signal-to-noise ratio of [value missing] was actively injected into the pose matrix of the virtual camera. Gaussian pose noise; the experimental group used the hierarchical loading method based on motion compensation weights of this invention, while the control group used a detail level switching method based on a single Euclidean distance; in the simulated virtual camera execution Under rotating conditions, the experimental group reduced the weight-sensitive angle in real time. to Limit the offset angle Greater than Geographic entity nodes are placed into a low-priority scheduling queue, while the control group maintains equal loading requests for all field-of-view nodes.

[0026] Table 1: Performance Comparison Data at Different Interaction Angular Velocities

[0027] Refer to the test results in Table 1, when the angular velocity achieve At that time, the large number of edge data loading requests generated by the control group crowded out the video memory bandwidth, causing the texture loading delay in the center region of the field of view to increase to [missing information]. And the frame rate dropped to The critical point; the experimental group of this invention dynamically adjusted the weight-sensitive angle. Make rendering load weight The distribution of textures is concentrated towards the center of the view axis, and its texture loading delay is... And the frame rate is stable at The above; when angular velocity Exceed Subsequently, the growth rate of texture loading latency tends to level off, and the determinism of rendering under motion conditions is ensured by locking the trend of weight evolution.

[0028] Example 3: This example combines Figures 1 to 2 This document describes a hierarchical loading and rendering method for three-dimensional geospatial data, such as... Figure 1 As shown, step S101 obtains the real-time pose matrix of the virtual camera, and extracts the virtual camera motion state parameters containing displacement vectors and rotational deflection amounts based on the real-time pose matrix. Step S102 calculates the weight-sensitive angle based on the numerical changes of the displacement vectors and rotational deflection amounts, and constructs a rendering weight distribution area centered on the view frustum. In step S103, the offset angle of the geographic entity node to be loaded relative to the view frustum is calculated, and the rendering loading weight of the geographic entity node is determined based on the proportional relationship between this offset angle and the weight-sensitive angle. Then, step S104 loads the rendering weights according to the weights. Geographic entity nodes are assigned to the corresponding asynchronous scheduling queue and matched with the level of detail to perform geospatial data parsing with the corresponding precision. Simultaneously, step S105 monitors the frame consumption of the rendering pipeline in real time. When the frame consumption exceeds the threshold, the level of detail of the geographic entity nodes in the asynchronous scheduling queue is downgraded in order of increasing rendering loading weight until the output frame rate reaches the stability threshold. Finally, step S106 calls the rendering shader to perform geometric feature linear interpolation on the geographic entity nodes in the process of switching levels of detail to smooth the geometric feature jumps during the switching process.

[0029] like Figure 2 As shown, the overall architecture starts with a virtual camera driver as the initial interaction point. This driver is used to extract motion state parameters and construct a rendering weight distribution area, serving as the input for determining the node rendering loading weight. The node rendering loading weight determination step outputs control instructions to the asynchronous data scheduling module, which is responsible for retrieving data from external storage media, adjusting the queue according to the weight, and establishing a connection with the hierarchical degradation processing module. In another logic, the rendering pipeline frame time monitoring module obtains time data from the rendering pipeline monitoring object. When the time exceeds the threshold, it triggers the execution of the hierarchical degradation processing module. The data flow of the hierarchical degradation processing module finally flows into the geometric feature linear interpolation step, which works with the graphics processor (GPU) and calls the shader to complete the final image rendering output.

[0030] Example 4: In a large-scale real-scene 3D geographic information system with a cloud-based asynchronous rendering architecture, the system faces challenges such as network bandwidth fluctuations and the presence of client-side sensors. Frequency jitter; if parameter determination procedures are lacking, the level of detail switches within the limited downlink bandwidth, inducing flickering of the model surface texture. The system executes the discretization path determination of the weight decay function: obtaining the offset angle of the geographic entity node relative to the view axis. and weight-sensitive angle and calculate and The ratio; the system adopts a weight evolution model based on trigonometric functions, and its calculation formula is: ,in, Weights are loaded for rendering. This represents the offset angle of the geographic entity node relative to the view axis. For weight-sensitive angles; when Approaching hour, Get the maximum value ;when equal hour, Value The system assigns geographic entity nodes to low-priority queues and converts the spatial distribution characteristics within the view frustum into quantified resource allocation indicators through mathematical monotonicity constraints.

[0031] The system establishes the admission criteria for asynchronous scheduling queues through a weighted threshold interval mapping procedure. The comparison is performed with preset priority thresholds to determine the level of detail allocation scheme; and the results are then... In to Nodes within a given interval are assigned to a high-priority queue, with a priority level no lower than [previous priority level]. The memory addressing bandwidth is matched to the highest level of detail, and the system will... In to The nodes of the interval are assigned to the medium-priority queue. The bandwidth resources are matched with the normal detail level, and the bandwidth resources are matched with the normal detail level. Below Nodes are assigned to low-priority queues, retaining only rendering requests for basic geometric skeletons; to address scheduling logic oscillations caused by sensor noise, the system executes a sliding window-based perception inertial calibration procedure, storing geographic entity nodes in the preceding sequence. Historical stable weights within the frame rendering cycle And calculate the current rendering load weight. Compared to The deviation of the mean is used to set the de-jitter gate threshold to . If the deviation is not greater than the de-jitter gating threshold, maintain the current scheduling state, and when the virtual camera's angular velocity... Continuous and continuous offset The deviation of the frames is greater than Update The value is obtained and the asynchronous scheduling process is restarted. The computational overhead caused by random jitter is filtered out through the integral smoothing effect in the time domain.

[0032] Example 5: In a 3D digital twin deployment environment with heterogeneous hardware terminals, a rendering pipeline benchmark performance calibration procedure is executed before system startup. The vertical synchronization signal period of the current device is captured using the graphics processor's internal timer, and the trigger step size for graded degradation processing is established accordingly. The idle time of three consecutive rendering frames is monitored in silent mode, and the average idle time per frame is calculated. Including screen refresh rate Together they determine the preset delay threshold The initial value is calculated using the following formula: ,in, The preset delay threshold is expressed in units of 1000 m / s. , This refers to the screen refresh rate, measured in units of... If the screen refresh rate for Then the system determines for The system allocates a grid addressing depth within the video memory buffer that matches the hardware performance level; in virtual reality software testing scenarios involving high-frequency rotational interactions, the system executes weight-sensitive angles. The discretized gradient filling procedure utilizes a visual tracking device to collect the user's perception and recognition rate of geographic entities at the edge of the field of view at different angular velocities, constructing a reference sequence for the evolution of the perceptual aperture. for to Within the range To extract measurement samples for the step size, when the perception recognition rate drops to a preset value... When the gating threshold is reached, the current critical angle value is recorded and stored in the perception lookup table to guide the decision-making on the level of detail degradation during dynamic interaction; correction is made by retrieving linear interpolation values ​​from the perception lookup table. The value of is determined to ensure that the perception evaluation model and the user's visual physiological characteristics are synchronized within the time interval.

[0033] When deployed on a high-performance virtual reality workstation with a distributed memory scheduling architecture, the system performs a baseline calibration procedure for memory bandwidth saturation before accessing the real-world 3D dataset, and uses the memory monitoring interface to obtain the bus throughput limit of the current graphics processing unit at the peak rendering time in real time. The maximum number of asynchronous scheduled tasks allowed within a single frame is determined by simulating high-concurrency texture retrieval requests; when multiple geographic entity nodes simultaneously trigger detail level switching and the total bandwidth requirement exceeds... At that time, the system determines the time based on the corresponding node. Values ​​are assigned normalized weights, where the values ​​in the high-priority queue are assigned the first value. Bandwidth resources allocated to each geographic entity node Satisfies the calculation formula: ,in, To be assigned to the The bandwidth resources of a geographic entity node, in units of , This is the upper limit of bus throughput, and its unit is . , For the first The rendering loading weight of each geographic entity node. Currently in an active scheduling state The sum of the weights of all geographic entity nodes; if the system detects If the throughput falls below the minimum threshold required for the corresponding level of detail, the task is automatically postponed to the next rendering cycle. In a virtual reality scenario simulating the inspection of an urban traffic hub, the system executes a timing calibration procedure for smooth transitions between levels of detail, using the timer of the rendering shader to control the blending factor. Real-time increments are used to establish the execution cycle of linear interpolation of geometric features; the single-frame rendering interval is obtained. Based on the preset level of detail, the evolution time is... Calculate the mixing factor The value of the mixing factor Satisfies the iterative formula: ,in, The blending factor for the current rendering cycle. The blending factor of the previous rendering cycle. This is the single-frame rendering interval, and its unit is . , The total duration of the detail level switching, in units of ;like for And the system determines for ,but The single-frame increment is This enables geographic entity nodes to be in Intra-frame geometric evolution is performed; weights are loaded during rendering. The value remains at the preset minimum weight value for an extended period of time. When the survival threshold is reached, the system sends a memory reclamation command and blocks the grid addressing port of the geographic entity node.

[0034] Example 6: In military or aviation simulation scenarios where rendering stability is extremely critical, the system executes lower-level hardware resource management procedures based on rendering load weights. When determining the level of detail for geographic entity nodes, a threshold-based discretization mapping is performed, and continuous floating-point rendering is weighted. Multiply by the preset total number of detail levels Perform floor operations to obtain the base level index The register is set to a value range of 1000. to Weighted dead zone threshold Rendering load weight in the current rendering cycle The absolute value of the difference between the cached historical weight value and the previous cycle's weight value is greater than the weight dead zone threshold. At that time, the processor sends an updated base level index to the rendering pipeline. The instruction uses numerical Schmitt trigger logic to filter out hierarchical index jumps caused by truncation errors in floating-point calculations.

[0035] The asynchronous scheduling queue is instantiated in system memory as a binary max-heap data structure based on array storage, and the weights are rendered and loaded. Directly mapped to the binary max-heap sort key value, the number of requests waiting to be processed in the asynchronous scheduling queue has reached the preset queue depth limit. When the weighted last-place elimination is executed, the texture pre-fetch task corresponding to the leaf node with the smallest sorted key value in the binary max-heap is removed. The released queue storage slots are then reassigned to newly entered geographic entity nodes with sorted key values ​​higher than the leaf node. The direct memory access controller only performs this action when it detects that the bus load rate is below a certain threshold. During the idle period, data loading instructions are extracted from the top position of the binary max-heap and submitted in batches to the video memory interface.

[0036] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A hierarchical loading and rendering method for three-dimensional geospatial data, characterized in that, Includes the following steps: Step S101: Obtain the real-time pose matrix of the virtual camera, and extract the motion state parameters of the virtual camera based on the real-time pose matrix. The motion state parameters include displacement vector and rotation deflection. Step S102: Based on the numerical changes of displacement vector and rotation deflection, calculate the weight-sensitive angle of the virtual camera in the current rendering cycle, and construct a rendering weight distribution area with the view axis as the center axis within the view frustum of the virtual camera. Step S103: Obtain the spatial location data of the geographic entity node to be loaded, calculate the offset angle of the geographic entity node relative to the view axis, and determine the rendering loading weight of the geographic entity node based on the ratio of the offset angle to the weight-sensitive angle. Step S104: Based on the rendering loading weight, the geographic entity nodes are assigned to the corresponding asynchronous scheduling queue, and the corresponding level of detail is matched according to the size of the rendering loading weight to perform geospatial data parsing with the corresponding precision. Step S105: Monitor the frame latency of the rendering pipeline in real time. When the frame latency exceeds a preset latency threshold, perform graded degradation processing on the detail level of the geographic entity nodes in the asynchronous scheduling queue according to the rendering loading weight from small to large, until the output frame rate of the rendering pipeline reaches a preset stability threshold, wherein the stability threshold is not lower than... ; Step S106: Invoke the rendering shader of the graphics processor to perform geometric feature linear interpolation on the geographic entity nodes that are in the process of switching levels of detail, so as to smooth the geometric feature jumps in the process of switching levels of detail.

2. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, In step S103, determining the rendering loading weight of a geographic entity node includes: comparing the offset angle with the weight-sensitive angle; if the offset angle is greater than the weight-sensitive angle, the geographic entity node is determined to be in a low-weight area, and the rendering loading weight is determined based on a preset minimum weight value; if the offset angle is less than or equal to the weight-sensitive angle, the rendering loading weight is calculated based on the ratio of the offset angle to the weight-sensitive angle using a preset weight decay function.

3. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, Step S101 extracts the motion state parameters of the virtual camera, including: calculating the pose transformation matrix between the current rendering cycle and the previous rendering cycle, and decomposing the pose transformation matrix to obtain the displacement vector representing spatial translation and the rotation deflection amount representing the change of the view axis rotation angle.

4. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, Step S105 performs a graded degradation process, including: identifying a set of non-core nodes whose rendering loading weight is lower than a preset weight threshold; sending a scheduling suspension instruction through an asynchronous scheduling queue to terminate asynchronous I / O requests for texture map data in the set of non-core nodes, and forcibly limiting the set of non-core nodes to only rendering the preset basic level of detail model.

5. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, In step S104, the geographic entity nodes are assigned to the corresponding asynchronous scheduling queues, including: dividing all geographic entity nodes within the view frustum into high-priority queues, medium-priority queues, and low-priority queues according to the rendering loading weights; and allocating system bus bandwidth resources and video memory addressing space to each queue in descending order of priority.

6. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, In step S106, geometric feature linear interpolation is performed, characterized in that the linear interpolation follows the following calculation rules: ,in, To obtain the geometric feature values ​​of geographic entity nodes after performing linear interpolation. These are the geometric feature values ​​of geographic entity nodes at a low level of detail. These are the geometric feature values ​​of geographic entity nodes at a high level of detail. This is a preset blending factor representing the timing progress of the switching of detail levels.

7. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, The calculation of the weight-sensitive angle in step S102 includes: obtaining the real-time angular velocity of the virtual camera; and when the real-time angular velocity increases, synchronously decreasing the value of the weight-sensitive angle so that the rendering weight distribution area shrinks towards the central axis region.

8. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, Also includes: A scheduling de-jitter step based on historical weights is introduced. This step involves storing the historical stable weights of geographic entity nodes within a preset number of preceding frames and calculating the real-time change rate of the rendering loading weights relative to the historical stable weights. If the real-time change rate is lower than a preset gating threshold, the current level of detail of the geographic entity node is maintained.

9. The hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, The time delay threshold in step S105 is limited to Frame latency is determined by monitoring the closed-loop latency of the rendering pipeline from task submission to frame buffer rendering.

10. A hierarchical loading and rendering method for three-dimensional geospatial data according to claim 1, characterized in that, Step S104 involves performing geospatial data parsing at the corresponding precision, including: retrieving the corresponding grid topology data and texture resources from the external storage medium according to the level of detail, and injecting the grid topology data and texture resources into the video memory buffer using a multi-threaded asynchronous I / O mechanism.

Images