水深現代化係以全球導航衛星系統(Global Navigation Satellite System, GNSS)與慣性導航系統(Inertial Navigation System, INS)結合多音束測深儀(multibeam echosounder, MBES)直接量測海床橢球高，再透過離距模型（Separation Model, SEP）轉換至海圖基準面，取代傳統利用潮位資料改正和提供海圖基準。此方法除能避免區域基準不一致，亦不需額外改正潮位變化、動態吃水等垂直因子，進一步減少誤差來源。搭載MBES之無人水面載具(Unmanned Surface Vehicle, USV)能於狹小和淺水環境中靈活作業，補足場域限制造成的資料缺失。然在港灣沿岸或橋梁下等訊號遮斷情境下，GNSS/INS定位解遭受干擾，導致垂直定位精度下降，水深成果恐無法符合IHO特等精度規範。本研究提出一套結合 GNSS/INS與隨機森林(Random Forest)的半自動異常偵測與改正流程，以減輕人工判讀負擔並提升資料修正一致性。實驗場域涵蓋高雄港內泊位區、橋梁區及出海口開放水域。本研究先測試不同後處理定位模式於泊位區及橋梁區之定位差異(含GNSS與GNSS/INS之精密單點定位(Precise Point Positioning,PPP)與後處理差分動態定位(Post-Processed Kinematic, PPK))。其中泊位區之定位差異不大(差值平均與標準偏差低於3公分)，而橋梁區之GNSS/INS定位解(含鬆耦合(Loosely Coupled, LC)與緊耦合(Tightly Coupled, TC))較能適應因衛星訊號遮蔽造成的定位影響，其中TC可於低衛星數量下(低於4顆)維持GNSS/INS的定位解輸出，而PPK利用差分觀測技術於短時間內收斂受影響之定位精度，基於上述優點選擇TC PPK作為後處理定位最優解。後續利用經海水面起伏(Heave)修正之TC PPK橢球高與自設潮位站資料，於移除各自平均值後，計算兩者間差值並以設定之垂直偏差門檻進行異常標記(以5公分為主，亦測試6~8公分的標記結果)，使用隨機森林(Random forest)執行偵測異常之訓練與預測。模型輸入特徵包含多項定位品質指標（如可用衛星數量、垂直精度衰減因子(Vertical Dilution of Precision, VDOP)、內部精度）和時間動態指標。透過欠採樣與成本敏感學習技術補正異常樣本不足問題，另以特徵相關性與重要性建立特徵與垂直偏差門檻交集之異常標記法，替代垂直偏差門檻異常標記法，藉此提升模型辨識力與解釋性。改正結果顯示，整體差值平均改正率於橋梁區前段與後段分別達 99.49% 與 99.79%，平均絕對誤差(Mean Absolute Error, MAE)改正率為85.41%與92.76%，均方根誤差(Root Mean Square Error, RMSE)改正率則為83.47%與90.55%。而於移除橋墩與堤岸牆面點雲之最終成果，橋梁區前段異常改正前後之水深網格差值平均與標準偏差分別改善41%與31%，後段區域分別改善25%與26%，顯示此法能有效提升遮蔽區水深成果精度。然本方法之限制為異常標記高度依賴自設潮位站資料，但潮位資料具不確定性，影響異常標記的精準度。另外，不同測區異常級距差異大(如泊位區與橋梁區)，模型泛化仍需透過更多場域測試與模型優化改善。

This study evaluates a Depth Modernization workflow integrating GNSS/INS positioning with MBES on USV in Kaohsiung Port. The aim is to improve ellipsoidal-height solutions under signal obstruction and reduce manual effort through semi-automatic anomaly detection and correction.
Post-processed modes (PPP, PPK of GNSS & GNSS/INS) were tested across berth, bridge, and open-water zones, with tightly coupled PPK proved most stable. Anomalies were labeled using tide-gauge differences and classified with Random Forest using under-sampling and cost-sensitive learning.
Results show anomalies concentrated in bridge sections, where corrections reduced MAE/RMSE by up to 85%/83% (front) and 93%/91% (back), also improved depth-grid differences by 41%/31% (front) and 25%/26% (back).The method enhances accuracy and consistency of Depth Modernization in complex ports while reducing workload.

內政部，水深測量作業規範，2020。
內政部，我國垂直基準轉換模式精進工作案第四期成果報告，2020a。
內政部，我國近岸平均海水面與海潮模式精進工作案第四期成果期末報告，2020b。
內政部，112-113 年測深無人船實務應用發展及線上精密單點定位服務系統模組精進工作案第4期成果報告書，2024。
史天元、薛憲文，高程現代化於海洋測量之實踐，《國土測繪與空間資訊》，第10卷，第1期，pp.21 - 36，2022。
呂建興、黃金維、藍文浩、王成機、郭重言，建置我國垂直基準轉換模式，《國土測繪與空間資訊》，第1期，pp. 37 - 64，2022。
李建鋒、王成機，水深測量之潮位修正方法，《國土測繪與空間資訊》，第10卷，第1期，pp.65 - 75，2022。
張宇翔，應用無人遙控載具瑜湖泊水深地形測量之研究-以基隆市情人湖水域為例，河海工程學系，國立臺灣海洋大學，2022。
Breiman, L. Random Forests. Machine Learning, 45, pp.5–32, 2001. https://doi.org/10.1023/A:1010950718922
Buda, M., Maki, A., Mazurowski, M. A. A systematic study of the class imbalance problem in convolutional neural networks,Neural Networks, 106, pp. 249-259,2018.https://doi.org/10.1016/j.neunet.2018.07.011
Chiang, K.W., INS/GPS Integration Using Neural Networks for Land Vehicular Navigation Applications. Department of Geomatics Engineering, the University of Calgary, Calgary, Canada , 2004. https://doi.org/10.11575/PRISM/23670
Crocetti, L. Schartner, M., Soja, B., Discontinuity Detection in GNSS Station Coordinate Time Series Using Machine Learning. Remote Sensing,13(19) ,3906, 2021. https://doi.org/10.3390/rs13193906
Drummond, C., Holte, R.C. C4.5, Class Imbalance, and Cost Sensitivity: Why Under-Sampling beats Over-Sampling , 2003.
Duong, T.T. Integration Strategies and Estimation Algorithms to Improve the Navigation Accuracy of Land-Based Mobile Mapping Systems (Doctoral dissertation). Department of Geomatics, National Cheng Kung University, Tainan, Taiwan ,2013.https://doi.org/10.1007/978-3-319-72347-1
Dinov, I.D., Data Science and Predictive Analytics Biomedical and Health Applications using R, University of Michigan–Ann Arbor, USA ,2018.
Devkota, B.P., Saunders, L., Dhakal, R., Kandel, L.N. GPS Spoofing Detection with a Random Forest Multiclass Classifier,2024 IEEE Military Communications Conference (MILCOM), Washington, DC, USA, pp.202-208,2024. https://doi.org/10.3390/s25134045
Falco, G., Pini, M., Marucco, G. Loose and Tight GNSS/INS Integrations: Comparison of Performance Assessed in Real Urban Scenarios. Sensors. 17(2) ,255, 2017. https://doi.org/10.3390/s17020255
Kumar, D., Ghorbanpour , A., Yen , K., Soltani , I. Design and Implementation of a Dual Uncrewed Surface Vessel Platform for Bathymetry Research Under High-Flow Conditions, in IEEE Access, 13, pp. 106634-106651, 2025.https://doi.org/10.48550/arXiv.2502.12539
Gharbia,A.Y.A., Gomaa,A., Saleh,M., Mousa,A.E., Abbas,I.A., Hassan, M.R . GNSS geodetic velocity prediction using ensemble tree models in Abu-Dabbab, Egypt, The Egyptian Journal of Remote Sensing and Space Sciences, 28(2),337-347,2025. https://doi.org/10.1016/j.ejrs.2025.05.008
Hodnesdal,H., Herman ,I., Birgit K.L., Lars K.N., Arne E.O., Stig Ø. Using Ellipsoid as Vertical Reference for Seabed Mapping Why? How? , 2006. https://legacy.iho.int/mtg_docs/com_wg/IHOTC/IHOTC7/Vertical_Reference_for_Seabed_Mapping.pdf
He, H., Garcia, E. A. "Learning from Imbalanced Data," in IEEE Transactions on Knowledge and Data Engineering, 21(9), pp. 1263-1284, Sept. , 2009. https://doi.org/10.1109/TKDE.2008.239
Hovem, J. M. Ray trace modeling of underwater sound propagation. In Beghi, M. G. (Ed.), modeling and measurement methods for acoustic waves and for acoustic microdevices, 2013. https://doi.org/10.5772/55935
Hwang,C.W., Kuo,C.Y., Shih,H.C., Huang,W.H., Lan, W.H. Depth modernization by integrating mean sea surface model, ocean tide model and precise ship positioning. Journal of Geodesy, 99-26(3), 2025. https://doi.org/10.1007/s00190-025-01949-3
International Hydrographic Organization. IHO Transfer Standard for Digital Hydrographic Data (S-57 Edition 3.1), International Hydrographic Organization, Monaco, 2000.
International Hydrographic Organization. Standards for Hydrographic Surveys (S-44 Edition 6.1.0), International Hydrographic Organization, Monaco, 2020.
International Hydrographic Organization. IHO Electronic Navigational Chart(ENC) Product Specification (S-101 Edition 1.1.0), International Hydrographic Organization, Monaco, 2022.
Japkowicz N, Stephen S. The class imbalance problem: A systematic study1. Intelligent Data Analysis. 6(5), 429-449, 2002. https://doi.org//10.3233/IDA-2002-6504
Jouybari, A., Bagherbandi, M., Nilfouroushan, F. Impact of GNSS signal outage on EOPs using forward Kalman filter and smoothing algorithm, The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLIII-B2-2022, 59-64, 2022. https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-59-2022
Jai,V. Usability of multibeam echosounder for wreck investigations using backscatter and water column data in shallow waters, The International Hydrographic Review, 29(2), 2023. https://journals.lib.unb.ca/index.php/ihr/article/view/34698
Lipton, Z,C., Elkan, C., Naryanaswamy, B. Optimal Thresholding of Classifiers to Maximize F1 Measure. Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2014. Lecture Notes in Computer Science(), vol 8725. Springer, Berlin, Heidelberg, 2014. https://doi.org/10.1007/978-3-662-44851-9_15
Lycourghiotis, S., Kariotou, F. The 'GPS/GNSS on boat' technique for the determination of the sea surface topography and geoid: A critical review, Coasts, 2(4), 323-340, 2022.https://doi.org/10.3390/coasts2040016
Liu, T.Y. Semi-Automatic Outlier Identification Applied to Point Clouds from Modernized Bathymetric Surveying using Unmanned Surface Vehicle with Multibeam Echosounder, National Cheng Kung University, Tainan, Taiwan ,2024.
Makar,A. Limitations of Multi-GNSS positioning of USV in area with high harbour infrastructure, Electronics, 12(3),697, 2023. https://doi.org/10.3390/electronics12030697
Moradi, M., Zhai, Z.M., Nielsen, A., Lai, Y.C., Random forests for detecting weak signals and extracting physical information: a case study of magnetic navigation, APL Machine Learning, 2(1), 2024. https://doi.org/10.1063/5.0189564
National Oceanic and Atmospheric Administration (NOAA). Hydrographic Surveys Specifications and Deliverables, Office of Coast Survey, Silver Spring, Maryland, 2019.
Özbey ,V.,Ergintav, S.,Tarı, E. GNSS Time Series Analysis with Machine Learning Algorithms: A Case Study for Anatolia, Remote Sensing, 16(17), 3309,2024. https://doi.org/10.3390/rs16173309
Pehlivan,H. A novel outlier detection method based on Bayesian change point analysis and Hampel identifier for GNSS coordinate time series, EURASIP Journal on Advances in Signal Processing,1~25, 2024. https://doi.org/10.1186/s13634-023-01097-w
Schisterman, E,F., Faraggi, D., Reiser, B., Hu ,J. Youden Index and the optimal threshold for markers with mass at zero. Stat Med. Jan 30;27(2):297-315 ,2008. https://pmc.ncbi.nlm.nih.gov/articles/PMC2749250/?utm_source=chatgpt.com
Stateczny, A., Specht, C., Specht, M., Brčić, D., Jugović, A., Widźgowski, S., Wiśniewska, M., Lewicka, O. Study on the positioning accuracy of GNSS/INS systems supported by DGPS and RTK receivers for hydrographic surveys, Energies, 14(21), 7413, 2021. https://doi.org/10.3390/en14217413
Wibowo, A., Purnama, S.R., Pratama C., Heliani L.S., David P.S., Wibowo S.T. Anomaly detection on displacement rates and deformation pattern features using tree-based algorithm in Japan and Indonesia, Geodesy and Geodynamics,14(2), 150-162,2023. https://doi.org/10.1016/j.geog.2022.07.003
Youden, W. J. Index for rating diagnostic tests, Cancer, 3(1), 32–35, 1950.https://doi.org/10.1002/1097-0142(1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3
Yuan, G., Shen, L., Zheng, W.S. A Hybrid Method of Combinatorial Search and Gradient Descent for Discrete Optimization ,2017. https://arxiv.org/abs/1706.06493