AI based data-driven point cloud optimization strategy in cloud computing framework for improving rendering performance of virtual city scenes
Abstract With the development of cloud computing framework, the visualization of city scenes is shifting to the 3D domain. However, the increase of scene complexity makes efficient rendering of virtual city scene an important challenge. To address the problem of poor rendering of virtual city scene,...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-07-01
|
| Series: | Discover Computing |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s10791-025-09646-7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract With the development of cloud computing framework, the visualization of city scenes is shifting to the 3D domain. However, the increase of scene complexity makes efficient rendering of virtual city scene an important challenge. To address the problem of poor rendering of virtual city scene, the paper proposes an Artificial Intelligence (AI) based bundle adjustment point cloud optimization model based on cross-entropy loss by combining with the multi-detail hierarchy technique, the rendering efficiency is improved while guaranteeing the high consistency between the 3D model and the real world in terms of geometry. Comparison of the proposed method and model of the study revealed that the F1 value and accuracy of the proposed algorithm was 0.87 and 93.9% respectively, which is better than the comparison algorithm. Comparison results of the cross-entropy loss-based bundle adjustment point cloud optimization model revealed that the output of the optimized point cloud model basically focused on − 1 and 1 values instead of fuzzy output and showed strong robustness to data containing outliers and noise. The virtual city scene rendering experiments revealed that the compression of texture data was reduced from the initial 82.5GB to 24.7GB. The optimized virtual city point cloud model was able to be fully loaded into memory and video memory at any observation point. Time and frame rate experiments revealed that for Scenario 6 with complex data, the cross-entropy loss-based bundle adjustment point cloud optimization model achieved an average frame rate of 495 frames per second. It was able to achieve a faster rendering speed while maintaining a lower amount of data. The outcomes indicates that this method can effectively enhance the efficient rendering performance of virtual city scene and provide useful references and lessons for researchers of virtual city scene, to promote the continuous development and progress of virtual city scene rendering technology. |
|---|---|
| ISSN: | 2948-2992 |