Daily Arxiv

This is a page that curates AI-related papers published worldwide.
All content here is summarized using Google Gemini and operated on a non-profit basis.
Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.

Empowering Bridge Digital Twins by Bridging the Data Gap with a Unified Synthesis Framework

Created by
  • Haebom

Author

Wang Wang, Mingyu Shi, Jun Jiang, Wenqian Ma, Chong Liu, Yasutaka Narazaki, Xuguang Wang

Outline

This paper proposes a method to utilize 3D point cloud technology for efficient inspection of bridges suffering from aging and deterioration. To overcome the inefficiency of existing manual inspection methods and the limitations of insufficient real data, we present a systematic framework to automatically generate complete 3D bridge data. The framework generates complete point clouds with component-level instance annotations, high-fidelity colors, and precise normal vectors, and simulates various physically realistic incomplete point clouds to support the training of segmentation and completion networks. Experimental results show that the PointNet++ model trained with synthetic data achieves 84.2% mIoU in real bridge semantic segmentation, and the fine-tuned KT-Net shows excellent performance in component completion. This provides an innovative methodology and a basic dataset for 3D visual analysis of bridge structures, which has important implications for the advancement of automated infrastructure management and maintenance.

Takeaways, Limitations

•
Takeaways:
◦
Contributes to improving the efficiency of bridge inspection using 3D point cloud technology.
◦
We present a synthetic data generation framework to address the problem of insufficient real data.
◦
The accuracy of actual bridge data analysis was improved by improving the performance of PointNet++ and KT-Net models.
◦
Provides a basic dataset that can contribute to the development of automated infrastructure management and maintenance technologies.
•
Limitations:
◦
Further validation of the generalization performance of the proposed synthetic data generation framework is needed.
◦
Research is needed on its applicability and limitations to various types of bridge structures.
◦
It is necessary to consider errors and noise that may occur during data acquisition in a real environment.
👍
Made with Slashpage