為避免河道砂石堆積導致水災並影響河川周邊安全，河川疏濬是目前梳理河川的主要方式。然而，河川過度疏濬易影響河川平衡，導致下游砂石補充不足引發災害，因此須對河川疏濬的過程進行監測。近年來，遙測科技的不斷提升，加上可長時間及大範圍的紀錄空間的特性，已有許多研究利用無人機及超高解析度衛星立體對影像產製的數值地表模型(Digital Surface Model, DSM)廣泛使用在不同的地形上。本研究使用法國Airbus公司的Pléiades 1A/1B超高解析度衛星影像產製DSM (〖DSM〗_pla)、無人飛行載具系統(Unmanned Aerial Vehicle, UAV)產製之DSM (〖DSM〗_uav)及地面高程測量所提供之獨立數據，對疏濬過程中的河床變化進行監測。將三組資料進行多時序遙測分析法套合比對，以可靠地偵測主要地形變化。然而，因影像拍攝設備及拍攝條件影響，三維資訊有系統誤差及潛在變形等問題。為此本研究使用DSMs與地面高程測量點擬合三階三維多項式曲面進行全局高程套合，再將點雲以迭代最近點算法(Iterative closest point, ICP)局部套合，減少潛在變形所產生的問題。在三階三維多項式曲面修正後，Pléiades與UAV相對精度為0.7公尺，Pléiades及UAV與主管機關提供的地面檢測，在河道斷面平均高程的相對精度分別為0.47–0.84公尺及0.51-0.76公尺。在局部套合後，顯著補償 〖DSM〗_pla 和 〖DSM〗_uav 之間的潛在局部變形，進一步提高相對高程精度。遙測方式可提供有效率的面狀地形，彌補地面測量空間密度不足問題並可初步估算疏濬土方，故建議針對各大型疏濬區可辦理遙測方式配合地面測量，而達到實際監測之目的。;River dredging is essential for preventing flooding by removing excess sediment. However, excessive upstream dredging can cause downstream hazards due to sediment depletion, requiring process monitoring. With the advancement in remote sensing and the capability for monitoring land changes with wide spatiotemporal coverage, many studies have utilized the digital surface model (DSM) derived by Unmanned Aircraft Vehicle (UAV) and the very high resolution (VHR) satellite stereo images for various terrains. This study employs DSMs produced by UAV images (DSM_uav) and Pléiades 1A/1B satellite imagery (DSM_pla) to monitor riverbed changes during dredging. DSM_pla, DSM_uav and field survey data are compared to verify accuracy using multi-temporal remote sensing analysis. Deformations and systematic errors may exist in DSMs due to different acquisition equipment and conditions. Therefore, a 3rd-order 3D polynomial surface is fitted for consistent vertical datum alignment on a global scale using DSMs and the field survey points. Then, the DSMs are converted into point clouds (PC) and the iterative closest point algorithm (ICP) is used for local registration to reduce the problems caused by systematic errors. The accuracy of DSM_pla is improved to 0.7 meters after the 3rd order 3D polynomial surface correction. Further, refinement through local fine alignment of the point clouds significantly reduced minor misalignments between the DSM_pla and the DSM_uav, enhancing the relative elevation accuracy even more. Remote sensing methods can efficiently provide detailed surface terrain data, compensating for insufficient spatial density in ground measurements and enabling initial estimation of excavation volumes. Therefore, it is recommended to combine remote sensing methods with ground surveys in large-scale excavation areas to achieve effective monitoring objectives.
