- 期刊
- OpenAccess
透地雷達於樹木根系探測之訊號分析及標準化
Ground Penetrating Radar (GPR) Signal Analysis and Standardization for Tree Root System Detection
摘要
樹木根系的健全影響著樹木的健康與生長，然而，樹木根系大多位於地底，不易調查。為避免傳統破壞性檢測可能對樹木造成的傷害，非破壞性檢測之「透地雷達」為樹木根系或樹木健康檢測工具之另一種選擇。本研究透過砂坑試驗，建立透地雷達訊號分析之標準作業流程與參數設定，再將此方法應用於溪頭人造針葉林，以網格狀測線施作並做資料分析。研究建立之訊號分析流程如下：（1）地面校正，以初達波位置為深度0；（2）計算介質波速；（3）樹根回波圖判別；（4）軌跡平滑，以雙曲線波型之明顯亮暗帶像素數判斷移動窗格寬度，進行移動平均；（5）增益處理，隨著深度進行訊號放大；（6）移位處理，將分散之半雙曲線回波特徵轉換為點狀特徵；（7）計算訊號振幅變化，並將振幅變化標準化為0-99的區間；（8）振幅變化與樹根尺寸相關性分析；（9）閾值篩選與設定；（10）核密度估算，以空間中訊號之機率分布描繪地底形貌。結果顯示，此流程可以順利捕捉樹木根系訊號，且在溪頭人造林網格狀測線試驗中，可將地下部反射訊號轉換成類似樹根之連續分布樣態呈現。未來我們將透過小規模破壞性檢測，驗證訊號分析結果，為林木生理、森林經營等領域之研究提供正確的地下部根系資訊。
並列摘要
The soundness of a tree's root system determines the health and growth of the tree. However, tree root systems are mostly underground making them difficult to investigate. To avoid damages caused by traditional destructive testing methods, a non-destructive method using ground penetrating radar (GPR) is an option for tree root system detection and tree health assessment. In this study, we conducted a sandbox experiment to develop standard operating procedures (SOPs) for GPR data analysis and parameter setting. Afterward, we tested and applied the SOPs in a coniferous forest plantation in Xitou using grid lines followed by data analyses. The SOPs we developed for signal analysis were: (1) ground level calibration by taking the initial wave position as depth 0 ; (2) calculating the velocity of the medium; (3) interpreting the radar echo diagram of the tree roots; (4) smoothing by counting the pixels of bright and dark areas in a hyperbola as the window width to produce moving averages; (5) amplifying signals with depth; (6) migration by transforming the semi-hyperbola characteristics into point features; (7) calculating the amplitude difference and scaling the difference from 0 to 99; (8) conducting a regression analysis between the amplitude difference and the cross-sectional area of the tree roots; (9) setting thresholds to detect signals; and (10) delineating the underground morphology by estimating the kernel density of the spatial probability distribution of signals. Results showed that the developed SOPs could successfully determine the roots' signals and transform the underground signals into a continuous distribution that resembled the shape of the investigated tree roots using grid lines in a Xitou forest plantation. In the future, we will validate our analytical results by carrying out small-scale destructive testing, so that for research in the disciplines of forest physiology and forest management, information of the underground tree root systems can be accurately provided.