空載光達率定與點雲匹配

Abstract

空載光達系統（或稱為空載雷射掃描），為一種主動式之遙測技術用以快速獲取大量離散點三維坐標。空載光達系統的運作，基本上可視為透過快速旋轉反射鏡的雷射測距儀。由於其潛在之技術應用，使其在精度評估、資料校正（registration）及系統率定等相關議題上，吸引許多學者投入研究。空載光達點雲資料的系統誤差，其形成的原因很多，但主要來自於組成空載雷射掃描系統的 三個子系統，亦即雷射測距系統（laser ranging system）、全球定位系統（GPS）以及導航系統（IMU）。本研究擬藉由系統率定、資料精度評估以及殘餘系統誤差消除等三個觀點，提出一套檢核空載光達資料精度的方法。

就系統率定而言，首先探討每一個軸角（boresight）率定參數，其參數特性對掃描精度之影像，以及率定的方法。本研究中，介紹二種目前商用空載光達系統的率定方法；另就其操作上之缺點，提出改進之建議，同時使用實際率定飛航資料進行評估此改進方法，是否可提高軸角率定參數精度。其次，為能驗證率定參數用於實際雷射掃描業務時之精度，避免內插原始雷射點雲資料，且冀望能同時評估重疊航帶間之平面與高程精度，本研究利用點雲匹配概念搜尋重疊航帶間之對應點，以評估資料精度。一種常用於三維面匹配（surface matching）之演算法－疊代最近點演算法（Iterative Closest Point, ICP）。使用兩組經地面參考資料檢驗精度等級不同之掃描資料測試，結果顯示，ICP不僅能作為重疊航帶間之資料精度評估工具，同時也可以解決搜尋重疊航帶間對應點（correspondence problem）之問題，而這個問題在評估重疊空載雷射航帶間之精度及後續系統誤差改正時，為非常重要的一項課題。

當經由面匹配確認資料存在系統誤差時，最後一個步驟乃是利用航帶平差的概念，消除殘存之系統誤差。由於部分測試航帶內缺乏可供辨識之地面控制點資料，本研究分別採用三維相似轉換（亦即七參數轉換）及三參數航帶平差法，嘗試修正殘存之航帶系統誤差。同時，使用之輸入觀測資料，即是由前一步驟利用面匹配所獲得之對應點資料。

其次，利用三參數航帶平差法亦獲得與三維相似轉換近似之結果。因此，三參數航帶平差法除再次確認以面匹配所獲得之對應點資料可用作航帶平差中之共軛點外，也證實本法可吸收大部分之高程系統誤差。最後，歸結上述之研究成果，提出空載光達資料精度評估與系統誤差校正流程。

Airborne laser scanning (ALS), also known as “airborne LiDAR”, is an active remote sensing technique to capture surface terrain. The system is based on laser distance measurement, combined with a scanning mirror mechanism. As the potential of ALS becomes more promising, issues related to accuracy assessment, registration and data calibration receive increasing attention. Systematic errors in point clouds acquired by ALS may occur for many reasons. Three components of a laser system, namely, position (GPS), navigation (IMU), and range (laser scanner system), are sources of systematic errors. This dissertation presents a complete framework on handling the systematic errors in addressing system calibration, systematic error validation and remaining systematic error recovery.

For system calibration, each boresight misalignment parameter is discussed to assess its impact on data accuracy and methodology of recovery. The schemes on boresight calibration solution used by two different commercial systems are introduced and the improvement on one of these approaches is proposed. The in-situ data set from a calibration flight is used to evaluate the improvement on the accuracy of misalignment parameters. A surface matching method, i.e. the ICP algorithm, is proposed, for the validation of the calibrated point clouds. In addition, the ICP algorithm provides the benefit of avoiding the need to interpolate the raw laser points, and evaluating the height as well as the planimetry offsets from overlapping laser strips. To evaluate the performance of the algorithm across different data quality level, two data sets are tested. The results reveal that the ICP algorithm can be used to both quantify the discrepancies from overlapping strips, and identify a solution regarding the correspondence problem.

The remaining systematic errors can be affirmed by using the proposed surface matching technique. Next, this research presents a strip adjustment procedure for the recovery of data with remaining systematic errors. Two methods are applied. The first one is the three-dimensional (3-D) similarity transformation, i.e. the seven-parameter transformation between two 3-D data sets. The second one is the strip adjustment using three parameters to adjust the laser strips when not enough ground reference points are available. Meanwhile, the corresponding points derived from ICP matching are used to form the observations to implement the adjustment.

The two proposed methods of strip adjustment confirm the following: (1) the corresponding points from ICP matching are sufficient to form the observations to implement adjustment; (2) the two methods can recover systematic error, especially on height. Analysis of the proposed solution on corresponding finding is then presented. Finally, a scheme on the accuracy assessment as well as remaining systematic errors recovery for ALS data is proposed.