張量分析應用於結合空載光達資料與地形圖重建建物模型的品質預估之研究

Tensor Analysis for Quality Prediction of Building Model Reconstruction by Integration of Airborne LiDAR Data and Topographic Information

林柏丞(Bo-Cheng Lin) ； 尤瑞哲(Rey-Jer You)

《航測及遙測學刊》 20卷3期 (2016/05) Pp. 151-169

https://doi.org/10.6574/JPRS.2016.20(3).1

摘要

本研究應用張量分析於光達點雲的特徵萃取，並針對資料融合空載光達資料與地形圖產製的建物模型進行品質預估。特徵萃取是從空載光達資料中萃取屋頂面區塊與屋脊線段，屋頂面區塊的萃取是利用張量投票法(Tensor Voting Method, TVM)推論每一個空載光達點隱含的幾何特徵資訊，並利用主特徵為種子點的區域成長法將具有平面特徵的點雲群聚在一起。屋脊線段則是利用已萃取的屋頂面區塊推論而得。此外，本文提出三個正規化的特徵強度指標以減少點雲數量對特徵辨識的影響。針對TVM 萃取平面區塊的成果，除了第一類型與第二類型錯誤之外，還新增碎形錯誤與識別能力兩項新指標來評估。在資料融合空載光達資料與地形圖產製建物模型的過程，本研究引入穩健權函式的最小二乘法來匹配空載光達資料的建物屋頂面邊緣點與建物屋頂的二維向量圖的輪廓線，使兩組資料轉換至相同的坐標系統。基於融合資料的不一致會造成建物模型錯誤重建，本研究提出利用空載光達資料中建物屋頂面邊緣點與地形圖中建物輪廓線的殘差張量分析進行品質預估，目的是偵測在現有資料品質之下可能被重建錯誤的模型。實驗結果顯示，本研究所提的品質預估指標不僅提高自動化模型重建的可靠性，且減少人工檢核成果的時間。

關鍵字

空載光達 ； 地形圖 ； 張量分析 ； 資料融合

並列摘要

This study first presents a novel method based on the tensor voting framework for extracting building features from airborne LiDAR data. For the extraction of roof patches, geometric features of LiDAR points are represented by a tensor field. A region-growing method with principal features is developed from the properties of eigenvalues and eigenvectors of the tensor field. The extraction of ridge lines is then inferred from the segmented roof patches. On top of that, three new indicators of the strength of features are proposed to reduce the effect of the number of points on feature identification. Furthermore, Type I errors, Type II errors, fragmentation and discernment are used as quality indicators to represent the effectiveness of the proposed method. Next, we present an algorithm to integrate the LiDAR data and topographic maps for 3D building model reconstruction and develop a quality prediction indicator by the residual tensor analysis. To reduce influence of errors while integrating, a robust least squares method is applied to register boundary points extracted from LiDAR data and building outlines obtained from topographic maps. After registration, a quality indicator based on the tensor analysis of residuals is derived in order to evaluate the correctness of the automatic building model reconstruction. Finally, an actual dataset demonstrates the automatic model reconstruction quality of the predictions. The results show that our method can achieve reliable results and save both time and expense on model reconstruction.

並列關鍵字

airborne LiDAR ； topographic map ； tensor analysis ； data fusion