多星種高解析度SAR影像結合深度學習之船隻自動偵測與應用介面開發研究
Other Title
AUTOMATED SHIP DETECTION AND INTERFACE DEVELOPMENT USING MULTI-SATELLITE HIGH-RESOLUTION SAR IMAGERY AND DEEP LEARNING
Journal
中國土木水利工程學刊
Journal Volume
37
Journal Issue
5
Start Page
357
End Page
362
ISSN
1015-5856
Date Issued
2025-09
Abstract
本研究針對海域監控中船隻目標之自動化辨識需求，提出結合多星種高空間解析度合成孔徑雷達（SAR）影像與深度學習技術之自動偵測方法，並開發具人機互動功能之使用者操作介面。研究整合TerraSAR-X、ICEYE與Capella等多源SAR影像，透過YOLOv8模型進行船隻位置與構型自動辨識，分析影像空間解析度與目標特徵間之關聯。實驗結果顯示，1 m空間解析度與VV偏極條件下模型辨識效能最佳，召回率與精確率可為100%。所開發之模組化操作介面支援影像上傳、範圍選取、遮罩疊圖、屬性匯出與人工修正，提升實務應用之效率與可靠性。整體成果驗證該方法具備良好之泛化能力與部署潛力，對我國於邊境監控與非法作業偵測等場景具有應用價值。
This study proposes an automated ship detection framework that integrates multi- satellite high-resolution Synthetic Aperture Radar (SAR) imagery with deep learning techniques to enhance maritime surveillance. Leveraging datasets from TerraSAR-X, ICEYE, and Capella, a YOLOv8-based model is employed to detect ship positions and structures under varying polarizations and spatial resolutions. Experimental results indicate that the model achieves optimal performance under 1-meter resolution and VV polarization, with both precision and recall reaching 100%. The developed modular user interface supports image uploading, area selection, mask overlay, attribute export, and manual correction, enhancing operational efficiency and reliability in practical applications. The overall results verify that the proposed method possesses good generalization ability and deployment potential, offering practical value.
Subjects
合成孔徑雷達（SAR）
船隻自動偵測
深度學習
多星種資料融合
Synthetic Aperture Radar (SAR)
automated ship detection
deep learning
multi-satellite data fusion
Type
journal article