基於幾何特徵保留的自適應點雲簡化
Other Title
Adaptive Point Cloud Simplification Based on Geometric Feature Preservation
Journal
航測及遙測學刊
Journal Volume
30
Journal Issue
1
Start Page
23
End Page
46
ISSN
1021-8661
Date Issued
2025-03
Author(s)
黃琳軒
Abstract
光學點雲應用於各領域，龐大數據影響存儲與運算效能，如何簡化點雲並保留幾何特徵成為關鍵。現有方法多依賴經驗法則，難適應不同區域特性。本研究改進了一種基於邊緣點、特徵點及非特徵點的點雲簡化方法，採用自適應鄰域大小保留特徵。方法上，先濾除雜訊，利用主成分分析法獲取曲率與熵建立拓樸結構，區分散亂與規則區域。於提取邊緣點後，計算特徵點重要性，對非特徵點減點，完成簡化與特徵保留。實驗顯示本方法在不同場景中能有效保留主要幾何特徵，然高簡化率下，細微特徵可能模糊。除此之外，未來需透過自動化參數數值選定提升效能。
The application of photogrammetry point clouds across various fields often faces challenges due to the massive data volume, which affects storage and computational efficiency. Simplifying point clouds while preserving geometric features has thus become a critical task. Existing methods largely rely on empirical rules, making them less adaptable to varying regional characteristics. This study improves upon a point cloud simplification method based on edge points, feature points, and non-feature points, incorporating an adaptive neighborhood size to preserve features. The approach begins with noise filtering, followed by using Principal Component Analysis (PCA) to derive curvature and entropy for constructing a topological structure that distinguishes between irregular and regular regions. After extracting edge points, the importance of feature points is assessed, and non-feature points are reduced to achieve simplification with feature retention. Experimental results show that the proposed method effectively preserves major geometric features across different scenarios. However, at high simplification rates, fine details may become blurred. In the future, performance could be further enhanced by implementing automated parameter setup.
Subjects
點雲簡化
自適應鄰域
特徵保留
主成分分析法
Point Cloud Simplification
Adaptive Neighborhood
Feature Preservation
Principal Component Analysis (PCA)
Type
journal article