Scholarlink

Data fusion has been employed to combine multiple point cloud datasets to generate more comprehensive 3D building models. However, discrepancies among datasets can result in unresolved holes, reducing model accuracy. This study aims to enhance data fusion techniques using the advanced Laplacian method, known for its multi-view 3D hole extraction capability, to address simple holes with an additional dataset. The fused 3D building model is derived from three-point cloud datasets, and the automated procedure identifies and restores holes through two main phases: 3D hole extraction and Laplacian data fusion. Python scripting facilitates the application of the Laplacian technique. Experiments on a small part of the 3D model showed that simple holes were covered more accurately by substitute points, with CloudCompare evaluations indicating a significant reduction in mean distance to the target: from 0.45 m to 0.29 m for ALS points and from 0.53 m to 0.31 m for image-based points. Future research will address complex holes, incorporating additional constraints and line extraction from image data to accurately reshape these intricate gaps.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Simple Geometrical Holes in Fused 3D Building Model Using the Laplacian Method

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Wahyu Marta Mutiarasari,
  • Alias Abdul Rahman

摘要

Data fusion has been employed to combine multiple point cloud datasets to generate more comprehensive 3D building models. However, discrepancies among datasets can result in unresolved holes, reducing model accuracy. This study aims to enhance data fusion techniques using the advanced Laplacian method, known for its multi-view 3D hole extraction capability, to address simple holes with an additional dataset. The fused 3D building model is derived from three-point cloud datasets, and the automated procedure identifies and restores holes through two main phases: 3D hole extraction and Laplacian data fusion. Python scripting facilitates the application of the Laplacian technique. Experiments on a small part of the 3D model showed that simple holes were covered more accurately by substitute points, with CloudCompare evaluations indicating a significant reduction in mean distance to the target: from 0.45 m to 0.29 m for ALS points and from 0.53 m to 0.31 m for image-based points. Future research will address complex holes, incorporating additional constraints and line extraction from image data to accurately reshape these intricate gaps.