以波形堆疊法進行空載波形光達資料之 地面微弱回波訊號萃取

Abstract

全波形光達可快速的獲得大量地面三維點坐標，並紀錄完整波形訊號。相較於離散點光達，全波形光達透過後處理，可萃取更精確且豐富的地表物資訊。以往利用波形進行地面點萃取，使用一維波形分析方法。但當地面點受地物遮蔽，產生微弱回波訊號時，此訊號接近背景雜訊值，難以使用一維波形分析方法萃取出此訊號且容易求解發散。光達為連續性掃描，鄰近波形間具有空間相關性，以往僅分析單一波形，缺乏考量波形間之空間相關性，以萃取地表物三維坐標。本研究欲考量鄰近波形間之空間關聯性及地表連續性，經波形對位與疊合，萃取微弱地面回波訊號。 本研究目的為結合一維波形分析方法及波形堆疊法進行地面微弱回波萃取，分析波形間空間關聯性。研究方法共分為三部分，首先，使用高斯分解法進行單一回波訊號分析，高斯分解法求解步驟為雜訊濾除、初始值給定及高斯擬合；接著，選取鄰近波形，並進行波形對位與疊合，增顯地面微弱回波訊號；最後，使用高斯分解法擬合堆疊波形，萃取微弱地面點訊號。研究中，就測區之植物覆蓋程度、波形間之取樣距離、及不同波形取樣數進行成果分析及視覺化展示。 使用資料分別為Leica ALS60、Riegl Q680i及Optech Pegasus，測區位置為台灣那瑪夏區。因Leica ALS60波形取樣數較多、紀錄資訊較豐富，紀錄長度約為38.5公尺，因此森林區萃取成果主要使用Leica ALS60之資料進行分析；Riegl Q680i及Optech Pegasus資料用於不同波形取用數分析。成果顯示，植物覆蓋程度影響資料穿透率，當穿透率越高時，萃取正確率越高且增加點數量越多。分析不同波形取樣距離，當取樣距離越大時，波形間空間關係越弱，則萃取正確率降低。最後，當波形取樣數減少時，可記錄之空間距離降低，當測區樹木高度太高時，單一波形無法同時記錄樹頂及地面資訊，則不適用於波形堆疊法。

Airborne lidar is an advanced technology which can obtain three-dimensional coordinates and intensity value of ground objects efficiently. As technology developed, comparing with discrete lidar, full-waveform lidar records entire backscattered signal. Waveform analysis is to extract more information using offline processing. Traditionally, the received waveform is analyzed by 1-D waveform analysis method. However, the weak laser pulse is produced because of the dense tree coverage and the return signal is closer to background noise. The weak laser pulse is undetected using traditionally 1-D waveform analysis method. To overcome the over parameterizations of the waveform analysis method, this research combines Gaussian decomposition and waveform stacking methods by considering the geospatial relationship between adjacent waveforms. The proposed methods contain three major works. First, we utilize Gaussian decomposition to analyze every single received waveform. Gaussian decomposition is used to extract waveform attributes including peak, echo width, amplitude, return number, etc. Data smoothing, initialization and Gaussian fitting are the three major steps in Gaussian decomposition. Second, considering geospatial relationship between sequential waveforms, we align and stack adjacent waveforms to the related location for augmenting the weak signal. Finally, the stacked waveform is analyzed by Gaussian decomposition method and the information of weak return of the ground is extracted. The experimental data are acquired by Leica ALS60, Optech Pegasus and Riegl Q680i. The test area is located in the middle part of Taiwan. The experimental result indicates that when considering the geospatial relationships, the proposed method extract the weak returns on the ground successfully. The correctness and increasing rate of the extracted ground point is related to the vegetated coverage such as complexity and the dense of ground points. The increasing of sampling distances of adjacent waveforms reduces the correctness. Moreover, the number of sampling also affects the recorded length of signal. More samples produce more reliable results.