車載光達系統之輻射正規化研究

Abstract

光達(Light Detection and Ranging, Lidar)為一主動式遙感探測系統，可迅速且大範圍的獲取目標物的三維空間資訊。光達系統提供的特徵資料包含幾何資訊及輻射資訊，輻射資訊主要被使用於點雲的視覺化及分類，而在應用前必須先經過校正，原因是此種資訊會受到目標物的材質種類與儀器和目標物間的幾何關係等影響，產生不同的反射強度值。隨著全波形光達的發展，透過密集的取樣將完整的波形記錄儲存，使得過去無法得知的電磁波能量得以透過推導背向散射截面積取得。 在過去許多研究已經對空載光達的輻射校正進行討論，已有一整套完整的流程，主要的概念是以模型導向方法將反射強度值轉換成具有物理意義背向散射截面積，相較於空載光達的輻射校正的完整，地面光達及車載光達的輻射校正則還在發展階段，由於地面光達反射強度值於近距離時受系統因素的影響，並不像空載光達資料滿足距離平方成反比的定律，因此無法透過光達公式進行模型導向改正。本研究目的為建立地面光達及車載光達的正規化流程，透過資料導向方法找出地面及車載光達輻射正規化模式。 研究方法分為兩個部分，分別為地面光達及車載光達的反射強度值正規化。地面光達使用的系統為RIEGL VZ-400，車載光達系統為RIEGL VMX-250。其中，地面光達的改正是依據距離與反射強度值的實驗成果為主，擬合距離效應方程式；車載光達則是由資料中挑選出訓練區做為擬合方程式的基準，並提出了距離與入射角正規化方程式。 成果部分包含視覺展示，並針對各個測站、掃描器及掃描帶重疊區進行反射強度值差異的統計分析，以及分類的應用。成果顯示經過正規化後，於重疊區的反射強度值差值均有明顯的下降，地面光達的改善幅度達56.37%，而車載光達約50%的改善幅度。

Light Detection and Ranging (LiDAR) is an active remote sensing system, which can acquire the 3D spatial coordinates of the target effectively. Lidar point clouds provide both geometric and radiometric information. Radiometric information usually applies in visualization of point clouds and it should be corrected for further application. The reason is that the radiometric information is influenced by the material of the target and the geometric between target and the sensor. With the development of Full-waveform LiDAR, the electromagnetic energy can be applied for radiometric correction using Lidar equation. There are many studies related to the airborne laser scanning (ALS) radiometric correction. The concept of radiometric correction is to convert the intensity to backscatter cross-section with physical meaning value by a model-driven approach. Compared to the ALS radiometric correction, the terrestrial laser scanning (TLS) and mobile laser scanning (MLS) radiometric correction is an on-going research topic. Due to the internal system factors of TLS data, the TLS intensity does not follow the inverse square law like ALS data. So that the TLS data cannot simply uses the model-driven approach by LiDAR equation. The purpose of this research is to establish the workflow of radiometric normalization for TLS and MLS. The proposed scheme is a data-driven approach and it contains two major parts, i.e. TLS and MLS radiometric corrections. The TLS experimental data are acquired by RIEGL VZ-400 and MLS experimental data are acquired by RIEGL VMX-250. In TLS section, the normalization model is based on range-intensity experiment and model fitting. In MLS section, the normalization model is obtained from crossroad area. We consider the range effect and incident angle effect respectively. The experiments show the improvement visually, calculate the intensity difference in overlap area between stations, scanners and strips, and classify the road surface. The experimental results indicate that the intensity difference in overlap area have reduced obviously. The improvement for TLS and MLS were 56.37% and 50%, respectively.