整合光達資料與高解析衛星影像於建物偵測

Fusion of Lidar Data and High Resolution Satellite Imagery for Building Detection

張智安(Tee-Ann Teo) ； 陳良健(Liang-Chien Chen)

《航測及遙測學刊》 10卷4期 (2005/12) Pp. 361-371

https://doi.org/10.6574/JPRS.2005.10(4).5

摘要

本研究之目的為整合光達資料與高解析衛星影像於建築物偵測。工作重點包括資料前處理、區塊分割及建物偵測等三個部分。資料前處理包括點雲分類、網格化及空間套合，研究中使用有理函數轉換模式配合地面控制點進行空間套合，將光達資料與高解析衛星影像轉換到一致的坐標系統中。在區塊分割中，針對網格化光達資料及正射化高解析衛星影像進行區域分割，產生高程及光譜性質相似的區塊。在建物偵測部份，以區塊為單元進行決策樹分類，決策樹分類所使用的分類準則包括：高程、光譜、紋理及形狀特徵。研究中使用新竹科學園區之Leica ALS40光達資料及QuickBird衛星影像進行測試，並使用一千分之一數值地藉圖做為檢核。實驗結果顯示，結合光達資料及高解析衛星影像於建物偵測之精度優於89%。

關鍵字

光達 ； 高解析衛星影像 ； 區塊分割 ； 決策樹分類 ； 建物偵測

並列摘要

This investigation is to integrate the Lidar data and high resolution satellite image for building detection. The proposed scheme includes preprocessing, segmentation, and classification. The preprocessing includes point cloud classification, rasterization, and spatial registration. We employ ground control points and rational polynomial coefficients in spatial registration. Thus, the two data sets are unified in the same object coordinate system. In the segmentation, we combine the raster form Lidar data and high resolution satellite orthoimages. The data with similar heights and spectral attributes are merged into a region. In the classification, we use the region-based decision classification to separate the building and non-building regions. The attributes considered in the classification include: elevation, spectral, texture and shape of regions. LIDAR data acquired by Leica ALS 40 and QuickBird satellite images were used in the validation. A 1/1000 scale topographic map was used as ground truth. The experiment results indicate that the building detection may exceed 89% of accuracy.

並列關鍵字

Lidar ； high resolution satellite image ； segmentation ； decision tree classification ； building detection