DSpace collection: 博碩士論文

ICME特徵及其對Forbush Decrease的影響;ICME Features and Their Effects on Forbush Decreases

title: ICME特徵及其對Forbush Decrease的影響;ICME Features and Their Effects on Forbush Decreases abstract: Forbush Decrease（FD）指的是觀測到宇宙射線通量發生快速下降、再緩慢恢復的現象，而ICME事件被認為是引發FD事件的原因之一，雖然已經有不少相關的研究，然而至今仍有許多FD相關物理機制尚未被了解。本研究參考許多目前已發表的研究成果，挑選1998年至2024年的62例完整觀測之ICME事件，分析其震波、鞘層與磁雲區段的物理參數對銀河宇宙射線（GCR）在FD事件中下降幅度A_FD的影響。根據WIND衛星的電漿與磁場資料，計算各區段磁場強度B、風速V、密度N、溫度T及電漿β值、以及計算磁雲磁場的時鐘角與錐角等參數。並搭配奧盧中子監測站計數率及京都世界地磁資料中心的SYM-H指數，分析其震波與鞘層和磁雲區間的物理參數、磁雲中的磁場旋轉情形、南向磁場的數值與持續時間、太陽週期對FD現象的影響。本篇研究結果中，鞘層在B、V、T與A_FD具正相關，磁雲則在B、V與A_FD具正相關，磁雲旋轉的時鐘角在不同A_FD區間有不同的分布情形，南向磁場持續時間與A_FD無明顯相關，另外每年FD事件數與太陽黑子數變化趨勢相近。本研究整理了一份橫跨較長時間的事件列表、並分析不同因素對FD事件的影響，供想研究FD相關內容的學者參考。;Forbush Decrease (FD) refers to the phenomenon whereby galactic cosmic ray flux is observed to rapidly decrease and then slowly recover. ICME events are considered one of the causes triggering FD events. Although considerable related research has been conducted， many physical mechanisms related to FD remain not yet understood. This study references numerous published research findings and selected 62 observed ICME events from 1998 to 2024， analyzing the influence of physical parameters in their shock， sheath， and magnetic cloud regions on the decrease amplitude A_FD of galactic cosmic rays (GCR) during FD events. Based on plasma and magnetic field data from the WIND spacecraft， we calculated magnetic field strength B， wind speed V， density N， temperature T， and plasma β values for each region， as well as the clock angle and cone angle parameters of the magnetic cloud magnetic field. Combined with the Oulu neutron monitor count rate monitoring station and SYM-H index from the Kyoto World Data Center for Geomagnetism， we analyzed the physical parameters of shock， sheath， and magnetic cloud regions， magnetic field rotation patterns within magnetic clouds， southward magnetic field values and duration， and the influence of solar cycle on FD phenomena. The research results show that in the sheath region， B， V， and T exhibit positive correlation with A_FD， while in the magnetic cloud， B and V exhibit positive correlation with A_FD. The clock angle of magnetic cloud rotation shows different distribution patterns in different A_FD intervals， southward magnetic field duration shows no obvious correlation with A_FD， and the annual number of FD events shows similar variation trends to sunspot numbers. This study compiled an event catalog spanning a relatively long time period and analyzed the influence of different factors on FD events， providing reference for scholars wishing to study FD-related content.

電離層天波越地平雷達系統建立模擬與比對;Development， Simulation， and Comparison of an Ionospheric Skywave Over-the-Horizon Radar System

title: 電離層天波越地平雷達系統建立模擬與比對;Development， Simulation， and Comparison of an Ionospheric Skywave Over-the-Horizon Radar System abstract: 高頻（High Frequency， HF）電波可經由電離層折射形成超視距天波傳播，其傳播路徑與可用頻率會受到電離層電子密度、Es layer、地磁場及水平電子密度變化等因素影響。本研究主要利用垂直與斜向電離層探測、高頻通訊實驗、固定與移動式遠距接收，以及三維射線追蹤方法，分析臺灣及周邊海域之高頻天波傳播特性，並探討由電離層觀測結果推估實際斜向傳播條件與通訊可行性。本研究首先利用中壢垂直 ionogram 與滿州斜向 ionogram，建立中壢—滿州約 309 km 固定斜向天波鏈路之 transmission curve 分析方法，推算反射虛高、反射點電漿頻率、入射角與地表傳播距離。結果顯示，不同時段與頻率下之平均距離誤差絕對值約為 0.01–2.48 km，相對誤差最高約 0.8%。將相同方法延伸至 659 km 傳播距離後，平均距離誤差絕對值約為 0.40–4.83 km，相對誤差約為 0.061%–0.733%，顯示 transmission curve 配合實測 ionogram 可有效推估固定斜向鏈路之地表傳播距離。此外，本研究分析中壢—滿州鏈路之日出瞬變現象，日出期間 F layer 反射虛高下降率約可達 83 km/hr，入射角增加率約可達 9°/hr。Es layer 傳播結果之平均地表距離為 311.5 km，標準差為 3.5 km，與實際 309 km 鏈路距離之平均相對誤差約為 0.78%。在實際高頻通訊部分，本研究建立中壢—東沙島約 659 km 天波通訊實驗鏈路，將 Barker code 加碼之 BPSK 數位訊號轉換為通訊機可處理之語音頻帶音訊訊號，再利用 H2010 高頻通訊機進行天波發射與接收。接收資料經同步、頻率偏移估測、相位補償及相關解碼後，未進行相位補償時之整體解碼成功率約為 5.6%–21.2%，加入相位補償後提升至約 64.7%–86.2%；進一步使用 PCA 與 PLL 進行相位校正後，最高成功率分別可達 88.9% 與 89.6%，顯示接收訊號之相位變化為影響實際天波數位通訊解碼的重要因素。本研究亦發現，增加 Barker code 重複次數雖可提高理論處理增益，但過長之訊文時間會受到電離層通道相位變化影響，使解碼效能反而下降。此外，本研究整合 H2010 sounding mode、MUF 預報、中壢 ionogram 及勵進號船載接收資料，分析固定與移動接收條件下之 HF 傳播特性，並建立由 sounding mode 斜向觀測結果轉換為等效垂直電離層參數之方法，使通訊機觀測資料可進一步與 ionogram 進行比較，作為分析岸基發射至遠距固定或移動接收端傳播條件之依據。最後，本研究建立三維 ray tracing 模擬流程，結合 IRI 電子密度模型與 IGRF 地磁場模型，並以 PHaRLAP 進行交叉驗證。於無水平電子密度梯度條件下，本研究與 PHaRLAP 所得到之射線落點與地表距離具有相近分布，顯示所建立之 ray tracing 流程可重現不同地磁及 O-wave、X-wave 條件下之傳播特性。進一步比較固定點對點地表距離估算結果，中壢—滿州 309 km 鏈路在無地磁場、O-wave 與 X-wave 三種 ray tracing 設定下之平均絕對距離偏差約為 8.73 km，而 transmission curve 結合實測 ionogram 後約為 0.70 km；659 km 傳播距離之兩者誤差則分別約為 37.17 km 與 1.58 km，ray tracing 之平均絕對距離偏差分別約為 transmission curve 方法的 12.4 倍與 23.5 倍。結果顯示，ray tracing 適合分析地磁場、射線型態與三維電子密度結構對傳播路徑之影響，而結合實際 ionogram 之 transmission curve 對固定點對點地表距離具有較穩定之估算結果。;High-frequency (HF) radio waves can propagate beyond the line of sight through refraction in the ionosphere. Their propagation paths and usable frequencies are affected by factors such as ionospheric electron density， sporadic E (Es) layers， the geomagnetic field， and horizontal variations in electron density. This study investigates HF skywave propagation over Taiwan and the surrounding seas using vertical and oblique ionospheric sounding， HF communication experiments， fixed and mobile remote reception， and three-dimensional ray tracing. The applicability of ionospheric observations to the estimation of practical oblique propagation conditions and HF communication feasibility is also examined. First， vertical ionograms observed at Zhongli and oblique ionograms received at Manzhou were used to establish a transmission-curve analysis method for the approximately 309-km fixed oblique skywave link between Zhongli and Manzhou. The virtual reflection height， plasma frequency at the reflection point， incidence angle， and ground propagation distance were estimated. The results show that the absolute values of the mean distance errors under different time periods and frequencies ranged from approximately 0.01 to 2.48 km， with a maximum relative error of approximately 0.8%. When the same method was extended to a propagation distance of 659 km， the absolute values of the mean distance errors ranged from approximately 0.40 to 4.83 km， corresponding to relative errors of approximately 0.061%–0.733%. These results demonstrate that the transmission-curve method combined with measured ionograms can effectively estimate the ground propagation distance of fixed oblique links. In addition， sunrise transient variations along the Zhongli–Manzhou link were analyzed. During sunrise， the F-layer virtual reflection height decreased at a rate of up to approximately 83 km/hr， while the incidence angle increased at a rate of up to approximately 9°/hr. For Es-layer-supported propagation， the mean estimated ground distance was 311.5 km with a standard deviation of 3.5 km， corresponding to a mean relative error of approximately 0.78% compared with the actual 309-km link distance. For the practical HF communication experiment， an approximately 659-km skywave communication link was established between Zhongli and Dongsha Island. Barker-code-encoded BPSK digital signals were converted into voice-band audio signals that could be processed by the HF transceiver and were then transmitted and received through H2010 HF transceivers. The received data were processed using synchronization， frequency-offset estimation， phase compensation， and correlation-based decoding. Without phase compensation， the overall decoding success rate was approximately 5.6%–21.2%. After phase compensation， the success rate increased to approximately 64.7%–86.2%. With further phase correction using principal component analysis (PCA) and a phase-locked loop (PLL)， the maximum success rates reached 88.9% and 89.6%， respectively. These results indicate that phase variations in the received signal are a major factor affecting the decoding performance of practical HF skywave digital communication. The results also show that， although increasing the number of Barker-code repetitions can theoretically provide greater processing gain， an excessively long message duration is more strongly affected by ionospheric channel phase variations and may consequently reduce decoding performance. In addition， H2010 sounding-mode observations， maximum usable frequency (MUF) predictions， Zhongli ionograms， and shipborne receiving data obtained onboard the R/V LEGEND were integrated to investigate HF propagation characteristics under fixed and mobile receiving conditions. A method was also established to convert oblique sounding-mode observations into equivalent vertical ionospheric parameters， allowing measurements obtained from the HF communication system to be further compared with ionograms. This provides a basis for analyzing propagation conditions from a shore-based transmitter to distant fixed or mobile receivers. Finally， a three-dimensional ray-tracing procedure was developed by combining the International Reference Ionosphere (IRI) electron-density model with the International Geomagnetic Reference Field (IGRF)， and the results were cross-validated against PHaRLAP. Under conditions without horizontal electron-density gradients， the ray landing points and ground-distance distributions obtained from the developed method were similar to those obtained from PHaRLAP， demonstrating that the developed ray-tracing procedure can reproduce propagation characteristics under different geomagnetic-field assumptions and O-wave and X-wave conditions. For fixed point-to-point ground-distance estimation， the mean absolute distance deviation of the Zhongli–Manzhou 309-km link under the three ray-tracing settings of no magnetic field， O-wave， and X-wave was approximately 8.73 km， whereas the transmission-curve method combined with measured ionograms yielded approximately 0.70 km. For the 659-km propagation distance， the corresponding deviations were approximately 37.17 km and 1.58 km. The mean absolute distance deviations obtained by ray tracing were approximately 12.4 and 23.5 times those obtained using the transmission-curve method for the 309-km and 659-km propagation distances， respectively. These results indicate that ray tracing is suitable for investigating the effects of the geomagnetic field， ray type， and three-dimensional electron-density structure on HF propagation paths， whereas the transmission-curve method combined with measured ionograms provides more stable estimates of ground distance for fixed point-to-point skywave links.

基於資料集感知評估之深度學習太空物件偵測：合成至真實遷移與邊緣部署分析;Dataset-Aware Evaluation of Deep Learning-Based Space Object Detection for Synthetic-to-Real Transfer and Edge Deployment

title: 基於資料集感知評估之深度學習太空物件偵測：合成至真實遷移與邊緣部署分析;Dataset-Aware Evaluation of Deep Learning-Based Space Object Detection for Synthetic-to-Real Transfer and Edge Deployment abstract: 隨著低地球軌道（Low Earth Orbit， LEO）衛星與太空碎片數量持續增加，太空態勢感知（Space Situational Awareness， SSA）之在軌目標偵測需求日益重要。太空光學影像具有目標尺度小、光照變化大，且訓練資料多仰賴模擬生成等特性，使模型效能不僅取決於架構本身，亦高度受資料分布、尺度組成與部署平台限制之影響。本研究因此以 dataset-aware 與 scale-aware 之觀點，以 YOLOv8n 為主要模型，於 SPARK2021、SPARK2022、SPARK2024 合成資料集及多種尺度重組（merged）版本下進行系統性評估，並於 SPARK2022 上與 RT-DETR 系列模型進行架構比較，以自行標註之 NASA真實影像進行 cross-domain evaluation，並將候選模型部署至 Jetson Orin Nano 以評估邊緣部署可行性。實驗結果顯示，合成域內準確率不足以作為模型選擇之依據：域內表現最佳之 SPARK2024 於 NASA 真實影像上之遷移表現最弱，以 SPARK2021 訓練之模型則取得最佳 NASA 表現。此差異可由訓練資料與目標域之尺度分布匹配程度解釋（Spearmanρ = −0.905）；且於未使用資料增廣之條件下，多尺度整合之 MS-MERGED 優於所有單一資料集，顯示當目標域分布未知時，多尺度資料整合為較穩健之預設訓練策略。資料增廣之效益亦非固定正向，而受訓練資料尺度分布調制：LD-MERGED 之增廣增益高達 +0.292，LD-MERGED-Cap 受控實驗證實限縮 large object 尺度範圍後，增廣差距即明顯縮小。模型架構方面，RT-DETR 於無增廣條件下展現較佳原生特徵建模能力，YOLOv8n 則於加入增廣後取得最高整體表現，並於 Jetson Orin Nano 邊緣部署中展現最佳之延遲、吞吐量與能源效率。綜合而言，本研究之貢獻不在於提出單一最佳訓練資料集，而在於建立一套能解釋並評估 synthetic-to-real 遷移表現之 dataset-aware 與 scale-aware 評估框架，並將模型選擇擴展至準確率、遷移能力與部署效率之三軸權衡，可作為未來光學 SSA 模型設計、跨域驗證與邊緣部署研究之參考。;With the rapid increase of satellites and space debris in low Earth orbit， space object detection has become an important task for Space Situational Awareness (SSA). Optical space imagery typically contains small targets， strong illumination variation， and relies heavily on synthetic training data， making detection performance dependent not only on model architecture but also on dataset distribution， object-scale composition， and deployment constraints. This thesis therefore presents a dataset-aware and scale-aware evaluation framework: YOLOv8n is systematically evaluated across the synthetic SPARK2021， SPARK2022， and SPARK2024 datasets and several scale-grouped merged variants， and compared against RT-DETR variants on SPARK2022; candidate models are then evaluated on a self-annotated NASA real-image test set and deployed on the Jetson Orin Nano. The results show that in-domain accuracy on synthetic data is an unreliable criterion for model selection: SPARK2024， the strongest in-domain setting， exhibits the weakest transfer to NASA real images， whereas models trained on SPARK2021 achieve the best NASA performance. This difference can be explained by the degree of match between the object-scale distributions of the training data and the target domain (Spearman $\rho$ = −0.905). Notably， without augmentation the multi-scale MS-MERGED setting outperforms all single datasets， indicating that multi-scale data integration is the more robust default strategy when the target-domain distribution is unknown. Augmentation benefit is likewise modulated by the scale distribution of the training data: LD-MERGED exhibits a gain as large as +0.292， and the LD-MERGED-Cap controlled experiment confirms that restricting the large-object scale range substantially narrows this gap. Architecturally， RT-DETR is stronger without augmentation， whereas YOLOv8n achieves the best overall performance once augmentation is applied and delivers the best latency， throughput， and energy efficiency in edge deployment. Overall， the contribution of this thesis lies not in identifying a single best training dataset， but in establishing an evaluation framework that explains and assesses synthetic-to-real transfer performance， extending model selection to a three-axis trade-off among accuracy， transferability， and deployability. The framework can serve as a reference for future optical SSA model design， cross-domain validation， and edge-deployment research.

應用深度學習於電離圖之回波結構辨識與特徵參數反演;Application of Deep Learning to Echo Structure Recognition and Feature Parameter Inversion in Ionograms

title: 應用深度學習於電離圖之回波結構辨識與特徵參數反演;Application of Deep Learning to Echo Structure Recognition and Feature Parameter Inversion in Ionograms abstract: 在過去的研究中，傳統電離圖（ionogram）判讀演算法若遇到高背景雜訊、強烈的射頻干擾（radio frequency interference， RFI）或複雜的多層結構，在提取電離層軌跡與參數標註時容易導致其精確性及穩定性降低。因此近年在電離圖判讀演算法開發中逐漸引入深度學習結合電腦視覺技術。在本研究中首先針對電離圖中高背景雜訊與射頻干擾的嚴重影響，在雷達資料預處理中引入恆虛警率（constant false alarm rate， CFAR）進行射頻干擾偵測，後續以三種射頻干擾抑制演算法進行抑制效果比較。結果證實自適應特徵子空間干擾抑制法為三者演算法當中表現最為優越，在極低的訊號與干擾加雜訊比（signal-to-interference-plus-noise ratio， SINR）的環境下，仍能將 SINR 大幅提升，並完整保留電離層回波功率。在深度學習於辨識電離圖結構中，本研究改良 U-Net 模型架構經由消融實驗（ablation study）確立最佳訓練組成。於測試集之像素語義分割（semantic segmentation）任務中，模型整體表現優良；在結構辨識任務中，模型在各類別具有非常優越的準確度。經由模型標註之虛高、臨界頻率與真實標籤之標註數值比較後亦具有極高的一致性。顯示即使在電離層軌跡功率微弱的情況下，模型仍能有效辨識整體的軌跡。與國際參考電離層模型（International Reference Ionosphere， IRI）對比後，臨界頻率兩者的表現相符，惟電離層探測儀反演之 F2 層峰高度範圍廣泛。後續進一步在地磁寧靜期下與福爾摩沙衛星七號（FORMOSAT-7/COSMIC-2， FS-7/C-2）無線電掩星（radio occultation， RO）資料進行對比，其 hmF2 相關係數較IRI模型高。本研究將電離層探測儀反演之地磁場強度與地面磁力計（magnetometer）之水平地磁場強度進行分析。在頻率域中，皆具有全日潮、半日潮以及八小時潮的明顯分量。而對電離層探測儀與地面磁力計之數據進行諧波擬合後，結果顯示兩者之間存在時間差。而此時間延遲不同步的原因可能為電離層中具有其他動力機制影響其結構。;In previous studies， traditional ionogram scaling algorithms often suffered from reduced accuracy and stability when extracting ionospheric traces and annotating parameters under conditions of high background noise， intense radio frequency interference (RFI)， or complex multi-layer structures. Consequently， recent developments in ionogram scaling algorithms have increasingly integrated deep learning with computer vision techniques. To address the severe impact of high background noise and RFI in ionograms， this study first introduces a constant false alarm rate (CFAR) detector in the radar data preprocessing stage for RFI detection， followed by a comparative analysis of three RFI mitigation algorithms. The results demonstrate that the adaptive eigen-subspace filtering for interference suppression (AEFIS) algorithm outperforms the other two methods. It significantly enhances the signal-to-interference-plus-noise ratio (SINR) even in environments with extremely low initial SINR， while completely preserving the ionospheric echo power. Regarding the deep learning-based recognition of ionogram structures， this study improves the U-Net architecture and establishes the optimal training configuration through an ablation study. In the pixel-level semantic segmentation task on the test set， the model delivers robust overall performance; in the event-level structure recognition task， the model achieves highly superior classification accuracy across all categories. Furthermore， the virtual height and critical frequency automatically annotated by the model exhibit exceptional consistency with the ground-truth values， demonstrating that the model can effectively recognize the complete traces even when the echo power is extremely weak. When compared with the International Reference Ionosphere (IRI) model， the critical frequency retrieved by the proposed system is in good agreement; however， the F2-layer peak height (hmF2) retrieved by the ionosonde exhibits a wider dynamic range. Further validation under geomagnetically quiet conditions against Formosat-7/COSMIC-2 (FS-7/C-2) radio occultation (RO) data reveals that the hmF2 retrieved by the ionosonde yields a higher correlation coefficient than that of the IRI model. Lastly， this study analyzes the geomagnetic field intensity retrieved by the ionosonde alongside the horizontal geomagnetic field intensity measured by ground-based magnetometers. In the frequency domain， both datasets exhibit prominent diurnal， semidiurnal， and terdiurnal tidal components. However， harmonic fitting of the ionosonde and magnetometer data reveals a distinct time lag between the two. This phase asynchronization suggests that other physical and dynamical mechanisms within the ionosphere actively modulate its vertical structure.