高頻岸基雷達原主要應用於海表流場與波浪觀測，惟其接收訊號亦包含船舶等海面目標之反射回波，因而具有發展為近岸廣域海事監測工具之潛力。然而，船舶回波常受到強烈一階布拉格海浪回波、背景雜訊、天線場型及距離衰減等因素影響，使實際可偵測範圍與定位可靠度仍缺乏完整的實海域量化評估。本研究以桃園北港站之低功率、開放架構16單元高頻陣列雷達為研究對象，建立一套由原始雷達訊號處理、船舶目標萃取、AIS時空配對至偵測效能與誤差診斷的實證分析架構，以評估既有高頻海流雷達應用於船舶監測之可行性及操作限制。 研究將連續雷達資料分割為較短之相干積分時段，以降低船舶移動造成的頻譜展寬，並排除一階布拉格回波所在頻帶，再透過陣列波束成形建立距離—方位角能量分布，利用空間峰值萃取船舶候選目標。偵測結果則與AIS船位、航速、航向及船舶尺度資料進行時空配對，進一步分析觀測距離、有效投影尺度、相對航向、天線增益、回波強度及平均波高對偵測表現的影響，並評估定位、航向及徑向速度之一致性。 結果顯示，北港站對船舶目標之實際有效偵測範圍約為25公里，且偵測能力具有明顯的距離與船舶尺度依賴性。惟此數值係於本站現行以海流觀測為主之發射參數，中心頻率26.77 MHz、低發射功率、380 kHz掃頻頻寬下所得，反映既有測流雷達在不更動硬體前提下之船舶偵測能力。距雷達10公里內之AIS匹配率約為50%至60%，10至20公里間則下降至約30%，超過20公里後匹配率快速降低。長度100公尺以下之小型船舶主要集中於15公里以內，而200公尺以上大型船舶可偵測至約25公里。回波強度整體隨距離增加而衰減；在相近距離下，船舶相對雷達之有效投影尺度愈大，通常具有較高回波能量與匹配率，其中距離5至10公里且投影尺度較大之船舶，局部匹配率可達70%以上。研究亦發現，觀測扇區兩側的回波與漏報分布並不對稱，顯示天線增益、局部電磁環境與場型遮蔽對偵測結果具有系統性影響。 在定位精度方面，樣本充足區域之距離誤差中位數多約為100至270公尺，且主要隨觀測距離及回波強度改變；相對航向本身並未呈現穩定的系統性影響。漏報案例主要集中於20至25公里及特定正方位角區域，反映船舶密度、訊號距離衰減與天線場型共同控制偵測效能。另有部分雷達連續軌跡無法與AIS資料配對，顯示系統可能具有補充合作式船舶通報資料之潛力，但此類目標仍可能包含AIS接收缺漏、錯誤配對或環境干擾，尚不能直接認定為非合作船舶。 綜合而言，本研究的主要貢獻在於建立高頻陣列雷達船舶偵測之實海域驗證及誤差診斷框架，並量化低功率雷達系統在不同距離、船舶尺度與觀測條件下的有效操作邊界。研究結果可作為後續改善天線場型、偵測門檻、目標追蹤、跨站整合及自動化品質控管之依據，亦顯示臺灣既有高頻海流雷達觀測網具有拓展至近岸海事監控與非合作目標輔助偵測的應用價值。;High-frequency (HF) ground-wave radar was originally developed primarily for observing ocean surface currents and waves; however， its received signals also contain reflected echoes from sea-surface targets such as vessels， giving it the potential to be developed into a tool for wide-area coastal maritime surveillance. Nevertheless， vessel echoes are frequently affected by strong first-order Bragg sea-echo， background noise， antenna patterns， and range attenuation， such that the actual detectable range and localization reliability still lack a comprehensive quantitative assessment in real sea conditions. This study focuses on a low-power， open-architecture 16-element HF phased-array radar at the Beigang station in Taoyuan， establishing an empirical analytical framework spanning raw radar signal processing， vessel target extraction， AIS spatiotemporal matching， and detection-performance and error diagnostics， in order to evaluate the feasibility and operational limitations of applying an existing HF ocean-current radar to vessel monitoring. The continuous radar data were segmented into shorter coherent integration intervals to reduce spectral broadening caused by vessel motion， and the frequency band occupied by first-order Bragg echoes was excluded. Array beamforming was then applied to construct a range–azimuth energy distribution， and vessel candidate targets were extracted via spatial peak detection. The detection results were spatiotemporally matched with AIS vessel positions， speeds， headings， and vessel-dimension data， allowing further analysis of the effects of observation range， effective projected dimension， relative heading， antenna gain， echo intensity， and mean wave height on detection performance， as well as an evaluation of the consistency of localization， heading， and radial velocity. The results show that the actual effective detection range of the Beigang station for vessel targets is approximately 25 km， but detection capability exhibits a pronounced dependence on range and vessel dimension. Within 10 km of the radar， the AIS matching rate is approximately 50%–60%， declining to about 30% between 10 and 20 km， and dropping rapidly beyond 20 km. Small vessels shorter than 100 m are concentrated mainly within 15 km， whereas large vessels exceeding 200 m can be detected out to approximately 25 km. Echo intensity generally attenuates with increasing range; at comparable ranges， vessels with a larger effective projected dimension relative to the radar typically exhibit higher echo energy and matching rates， with vessels at 5–10 km and larger projected dimensions achieving local matching rates exceeding 70%. The study also finds that the distributions of echoes and missed detections on the two sides of the observation sector are asymmetric， indicating that antenna gain， the local electromagnetic environment， and pattern shadowing exert a systematic influence on detection outcomes. Regarding localization accuracy， the median range error in regions with sufficient samples is mostly about 100–270 m and varies primarily with observation range and echo intensity， whereas relative heading itself shows no stable systematic effect. Missed-detection cases are concentrated mainly at 20–25 km and within certain due-azimuth regions， reflecting the joint control of detection performance by vessel density， signal range attenuation， and antenna pattern. In addition， some continuous radar tracks could not be matched with AIS data， suggesting that the system may have the potential to supplement cooperative vessel-reporting data; however， such targets may still involve AIS reception gaps， mismatches， or environmental interference， and thus cannot be directly identified as non-cooperative vessels. In summary， the principal contribution of this study lies in establishing a real-sea validation and error-diagnosis framework for vessel detection using HF phased-array radar， and in quantifying the effective operational boundaries of a low-power radar system under varying ranges， vessel dimensions， and observation conditions. The findings can serve as a basis for subsequent improvements in antenna pattern， detection thresholds， target tracking， cross-station integration， and automated quality control， and further demonstrate that Taiwan′s existing HF ocean-current radar observation network holds application value for extension to coastal maritime surveillance and the auxiliary detection of non-cooperative targets.
