在無線感測器網路中，由於感測器電池的不可替換性，有效的能源管理是一項重要的研究議題。既然通訊及偵測都會消耗感測器的能量，減少多餘偵測範圍的重疊，及降低重覆資料(duplicate data)的影響，可有效節省能量，延長網路生命週期。於本研究中，我們提出VERA (Voronoi dEtection Range Adjustment)，利用分散式Voronoi diagram演算法劃分各感測器負責監控的區域，並利用基因演算法計算每個感測器最合適的偵測範圍以節省能量，延長網路生命週期。此外，我們亦考慮偵測能力的限制，在減少感測器偵測範圍重疊的同時，也避免某些區域的偵測能力低於門檻值。在實驗模擬的部份，我們利用模擬系統驗證所提出的方法是否能有效降低各感測器偵測範圍的重疊性，並因偵測範圍降低而導致duplicate data的減少和整個感測器網路總能量耗損的減少。末了，也將驗證本方法是否能延長無線感測器網路的生命週期和達到滿足偵測機率的最低保證。 / In the wireless sensor networks, the batteries are not replaceable, efficient power management thus becomes an important research issue. Since both communication and detection consume energy, if we can largely decrease the overlaps among detection ranges and reduce the duplicate data then we can save the energy effectively. This will thus prolong the network lifetime. In this research, we propose a Voronoi dEtection Range Adjustment (VERA) method that utilizes distributed Voronoi diagram to delimit the responsible area for each sensor, and utilize Genetic Algorithm to compute the most suitable detection range for each sensor. As we try to decrease the detection ranges, we still guarantee to meet the lower bound of the sensor detection probability. Simulations showed that our method can decrease the redundant overlaps among detection ranges, minimize energy consumption, and prolong the lifetime of the whole network effectively.

防火牆是現今網路中的重要設備，負責區隔內部網路和公共網路，維護內部網路安全。然而防火牆也存在幾個重要的問題，首先，防火牆的規則是由網管人員設定，近年來隨著網路科技蓬勃發展、虛擬技術大量應用，此項工作已帶給網管人員龐大的負擔。其次，防火牆雖可隔離外部網路，阻擋有害流量，但對內部網路的防範卻毫無用武之地。目前市面上普遍使用入侵偵測系統(IDS)進行偵測，但僅能在發現攻擊行為後發出警告訊息，無法即時處理。最後，企業在連外網路部分，通常採用多條線路進行備援，並倚賴多路頻寬負載平衡器(Multi-WAN load balancer)增加頻寬的使用率，但在線路數量上卻受限於廠商所制定之規格，無法彈性調整。而在負載平衡演算法方面，也只能基於網路特徵(IP位置)、權重比例(weight)或是輪詢機制(round robin)，無法依據目前網路狀況做出更好判斷。 為改善上述問題，本論文在軟體定義網路(SDN)環境下，使用交換機取代傳統防火牆設備，透過封包分析與信任觀測區間達到規則學習，並整合Snort入侵偵測系統，透過特徵比對，找出危害網路環境之封包，即時阻擋該危險流量。本論文也提出基於隨需(on demand)概念，動態調整防火牆規則，降低管理人員負擔。最後利用交換機擁有多個實體通訊埠的概念 ，依需求可自由調整對外及對內線路數量，不再受限於廠商規格，取代傳統多路寬頻負載平衡器，建構更彈性的架構。並透過收集交換機上的實體埠與資料流表中的資訊，即時評估網路狀況，加強負載平衡。為驗證本論文所提出之⽅法的有效性，我們使用Linux伺服器架設KVM、OpenvSwitch以及POX控制器實際建構SDN網路環境，透過發送封包對防火牆提出請求，以驗證實驗方法的正確性。 根據實驗結果顯示，本論文所提出之概念均能正確運作，有效降低調整防火牆所需之人工作業。在多路寬頻負載平衡器部分，本研究所提出之負載平衡方法，與round robin負載平衡方法相較之下，在最佳情況下，能有效提升約25%平均頻寬使用率，並降低約17.5%封包遺失率。 / Firewall is an important device that is responsible for securing internal network by separating Internet from Intranet, but here are several existing issues about the firewall. First, the firewall rules are set by the network admistrator manually. Along with the vigorous development of Internet technologies and great amount of applications of virtual technology in recent years. This work burdens the network adminstrator with a heavy workload. Second, the firewall is able to isolate the external network from harmful traffic, however, it can do nothing to the internal network. The common situation is to use IDS to detect the harmful packet, but it can only send an alert message to the adminstrater, no more actions can be done. Finally, most companies use several ISP connections to assure fault tolerance and use Multi-WAN load balancer to integrate those connections to enhance bandwidth utilization. But the number of WAN/LAN ports is set by the manufacturer, and the load balance algorithm is also limited by the manufacturer. It offers only a few algorithms (network-based features, round-robin, etc.), and there is no other way to provide more efficient algorithms. In order to resolve the mentioned problems, we propose an automatic firewall based Software Defined Network (SDN). We use Openflow switches to replace traditional firewalls, the system is able to learn the rules automaticlly by packet analysis during an observation interval. We aslo integrate Snort Intrusion Detection System (IDS) to localize the dangerous packets and block them immediately. Next, we propose an on-demand based dynamic firewall rules adjustment mechanism which is able to reduce management workload. Finally, we implement a Multi-WAN load balancer architecture and provide a more efficient load balance algorithm by collecting port usage and firewall rule information. In order to verify the proposed methods, we implement a SDN environment by using Linux Ubuntu servers with KVM, Open vSwitch and POX controller. According to the experiment result, it proves that the proposed method is able to reduce the firewall configuration effectively. In the Multi-WAN load balancer, experiment results show that our method outperforms round-robin argrithom in terms of average bandwidth utilization and packet loss rate by 25% and 17.5%, respectively.

近年來科技、網路以及儲存媒介的發達，產生的資料量呈現爆炸性的成長，也宣告了巨量資料時代的來臨。擁有巨量資料代表了不必再依靠傳統抽樣的方式來蒐集資料，分析數據也不再有資料收集不足以致於無法代表母題的限制。突破傳統的限制後，巨量資料的精隨在於如何從中找出有價值的資訊。 以擁有大量輿論和人際互動資訊的社群網站為例，就有相關學者研究其情緒與股價具有正相關性，本研究也試著利用同樣具有巨量資料特性的網路新聞，抓取中央新聞社2013年7月至2014年5月之經濟類新聞共計30,879篇，結合新聞主題偵測與追蹤技術及情感分析，利用新聞事件相似的概念，透過連結匯聚成網絡並且分析新聞的情緒和股價指數的關係。 研究結果顯示，新聞事件間可以連結成一特定新聞主題，且能在龐大的網絡中找出不同的新聞主題，並透過新聞主題之連結產生新聞主題脈絡。對此提供一種新的方式來迅速了解巨量新聞內容，也能有效的回溯新聞主題及新聞事件。 在新聞情緒和股價指數方面，研究發現新聞情緒影響了股價指數之波動，其相關係數達到0.733562；且藉由情緒與心理線及買賣意願指標之比較，顯示新聞的情緒具有一定的程度能夠成為股價判斷之參考依據。 / In recent years, the technology, network, and storage media developed, the amount of generated data with the explosive growth, and also declared the new era of big data. Having big data let us no longer rely on the traditional sample ways to collect data, and no longer have the issue that could not represent the population which caused by the inadequate data collection. Once we break the limitations, the main spirit of big data is how to find out the valuable information in big data. For example, the social network sites (SNS) have a lot of public opinions and interpersonal information, and scholars have founded that the emotions in SNS have a positive correlation with stock prices. Therefore, the thesis tried to focus on the news which have the same characteristic of big data, using the web crawl to catch total of 30,879 economics news articles form the Central News Agency, furthermore, took the “Topic Detection & Tracking” and “Sentiment Analysis” technology on these articles. Finally, based on the concept of the similarity between news articles, through the links converging networks and analyze the relevant between news sentiment and stock prices. The results shows that news events can be linked to specific news topics, identify different news topics in a large network, and form the news topic context by linked news topics together. The thesis provides a new way to quickly understand the huge amount of news, and backtracking news topics and news event with effective. In the aspect of news sentiment and stock prices, the results shows that the news sentiments impact the fluctuations of stock prices, and the correlation coefficient is 0.733562. By comparing the emotion with psychological lines & trading willingness indicators, the emotion is better than the two indicators in the stock prices determination.

大地起伏值為正高與橢球高的差異量，如果取得高精度的大地起伏值，可以利用衛星定位測量施測橢球高並計算得到高精度的正高，其成本低廉，可望取代傳統的水準測量。而大地起伏值可以分為幾何法或重力法的大地起伏值，其中幾何法的大地起伏值計算方法簡易且精度高，可以利用曲面擬合方法取得之。但是幾何法的大地起伏值會受到地形起伏的影響，大範圍的曲面擬合會降低其精度。台灣的地形起伏大，難以進行大範圍曲面擬合。 於是本研究利用環域方法搜尋待測點位鄰近的水準點參與曲面方程式擬合大地起伏，試圖找到最適合的大地起伏擬合範圍。成果顯示：環域的範圍從10公里至30公里，利用二次曲面方程式擬合大地起伏在台灣平地區域能夠達到預測精度與內部精度同時低於5公分。另外由於衛星定位測量橢球高的誤差較高，需進行資料品質評估並進行粗差偵測。針對粗差偵測提出新的方法，利用最佳化演算法中的量子行為粒子群演算法計算最小二乘平差法中的權矩陣，期望能夠將粗差觀測量的權重降低，達到粗差偵測的效果。成果顯示最佳化權矩陣演算法，能夠將粗差對平差系統的影響量降到最低。 本研究建立一套台灣地區的大地起伏擬合作業程序：利用環域搜尋鄰近水準點、曲面方程式及環域範圍選擇與資料的粗差偵測，可獲得高品質的大地起伏。 / The geoidal undulation is the difference of ellipsoid height and orthometric height. We can obtain high accuracy of orthometric height by existing high accuracy of geoidal undulation and the ellipsoidal height measuring by GPS. It expected to replace the traditional leveling survey due to the less cost. This study uses buffer method to search the leveling benchmarks around the object point, attempts to find the proper range of fitting geoidal undulation to curve surface. Experimental results shows that it can archive 5cm level on both prediction error and internal precision by fitting geoidal undulation on 2nd curve surface model where the buffer range is from 10 km to 30 km. In this study, also uses the quantum-behaved particle swarm optimization to calculate the weight matrix of least square adjustment, the purpose is to down-weighting the suspicious outlier, and detect the outlier. Experimental results shows that the optimal weight matrix algorithm can reduce the influence of outlier. This study establish a procedure of fitting geoidal undulation: using buffer analysis to search the adjacent leveling benchmark, selecting the proper buffer range and surface equation and detecting outlier in data.

影像中的文字訊息，通常包含著與場景內容相關的重要資訊，如地點、名稱、指示、警告等，因此如何有效地在影像中擷取文字區塊，進而解讀其意義，成為近來電腦視覺領域中相當受矚目的議題。然而在眾多的場景文字偵測方法裡，絕大多數是以英文為偵測目標語言，中文方面的研究相當稀少，而且辨識率遠不及英文。因此，本論文提出以中文和英文為偵測目標語言的方法，分成以下四個主要程序：一、前處理，利用雙邊濾波器(Bilateral filter)使文字區域更加穩定；二、候選文字資訊擷取，考慮文字特徵，選用Canny 邊緣偵測和最大穩定極值區域(Maximally Stable Extremal Region)，分別提取文字邊緣和區域特徵，並結合兩者來優化擷取的資訊；三、文字連結，依中文字結構和直式、橫式兩種書寫方向，設置幾何條件連結候選文字字串；四、候選字串分類，以SVM加入影像中文字的特徵，分類文字字串和非文字字串。使得此方法可以偵測中文和英文兩種語言，並且達到不錯的辨識效果。 / Text messages in an image usually contain useful information related to the scene, such as location, name, direction and warning. As such, robust and efficient scene text detection has gained increasing attention in the area of computer vision recently. However, most existing scene text detection methods are devised to process Latin-based languages. For the few researches that reported the investigation of Chinese text, the detection rate was inferior to the result for English. In this thesis, we propose a multilingual scene text detection algorithm for both Chinese and English. The method comprises of four stages: 1. Preprocessing by bilateral filter to make the text region more stable. 2. Extracting candidate text edge and region using Canny edge detector and Maximally Stable Extremal Region (MSER) respectively. Then combine these two features to achieve more robust results. 3. Linking candidate characters: considering both horizontal and vertical direction, character candidates are clustered into text candidates by using geometrical constraints. 4. Classifying candidate texts using support vector machine (SVM), the text and non-text areas are separated. Experimental results show that the proposed method detects both Chinese and English texts, and achieve satisfactory performance compared to those approaches designed only for English detection.

近幾年在金融風暴及全球競爭等影響下，企業紛紛導入商業智慧平台，提供管理階層可簡易且快速的分析各種可量化管理的關鍵指標。但在後續的推廣上，經常會因商業智慧系統提供的資訊過於豐富，造成使用者在學習階段無法有效的取得所需資訊，導致商業智慧無法發揮預期效果。本論文以使用者在商業智慧平台上的操作相似度進行分析，建立相對於實體部門的凝聚子群，且用中心性計算各節點的關聯加權，整合至所設計的推薦機制，用以提升商業智慧平台成功導入的機率。經模擬實驗的證實，在推薦機制中考慮此因素會較原始的推薦機制擁有更高的精確度。 / In recent years, enterprises are facing financial turmoil, global competition, and shortened business cycle. Under these influences, enterprises usually implement the Business Intelligence platform to help managers get the key indicators of business management quickly and easily. In the promotion stage of such Business Intelligence platforms, users usually give up using the system due to huge amount of information provided by the BI platform. They cannot intuitively obtain the required information in the early stage when they use the system. In this study, we analyze the similarity of users’ operations on the BI platform and try to establish cohesive subgroups in the corresponding organization. In addition, we also integrate the associated weighting factor calculated from the centrality measures into the recommendation mechanism to increase the probability of successful uses of BI platform. From our simulation experiments, we find that the recommendation accuracies are higher when we add the clustering result and the associated weighting factor into the recommendation mechanism.

過往對於情緒分析的研究顯少聚焦在讀者情緒，往往著眼於筆者情緒之研究。讀者情緒是指讀者閱讀文章後產生之情緒感受。然而相同一篇文章可能會引起讀者多種情緒反應，甚至產生與筆者迥異之情緒感受，也突顯其讀者情緒分析存在更複雜的問題。本研究之目的在於辨識讀者閱讀文章後之切確情緒，而文件分類的方法能有效地應用於讀者情緒偵測的研究，除了能辨識出正確的讀者情緒之外，並且能保留讀者情緒文件之相關內容。然而，目前的資訊檢索系統仍缺乏對隱含情緒之文件有效的辨識能力，特別是對於讀者情緒的辨識。除此之外，基於機器學習的方法難以讓人類理解，也很難查明辨識失敗的原因，進而無法了解何種文章引發讀者切確的情緒感受。有鑑於此，本研究提出一套基於語意框架(frame-based approach, FBA)之讀者情緒偵測研究的方法，FBA能模擬人類閱讀文章的方式外，並且可以有效地建構讀者情緒之基礎知識，以形成讀者情緒的知識庫。FBA具備高自動化抽取語意概念的基礎知識，除了利用語法結構的特徵，我們進一步考量周邊語境和語義關聯，將相似的知識整合成具有鑑別力之語意框架，並且透過序列比對(sequence alignment)的方式進行讀者情緒文件之匹配。經實驗結果顯示證明，本研究方法能有效地運用於讀者情緒偵測之相關研究。 / Previous studies on emotion classification mainly focus on the writer's emotional state. By contrast, this research emphasizes emotion detection from the readers' perspective. The classification of documents into reader-emotion categories can be applied in several ways, and one of the applications is to retain only the documents that cause desired emotions for enabling users to retrieve documents that contain relevant contents and at the same time instill proper emotions. However, current IR systems lack of ability to discern emotion within texts, reader-emotion has yet to achieve comparable performance. Moreover, the pervious machine learning-based approaches are generally not human understandable, thereby, it is difficult to pinpoint the reason for recognition failures and understand what emotions do articles trigger in their readers. We propose a flexible semantic frame-based approach (FBA) for reader's emotion detection that simulates such process in human perception. FBA is a highly automated process that incorporates various knowledge sources to learn semantic frames that characterize an emotion and is comprehensible for humans from raw text. Generated frames are adopted to predict readers' emotion through an alignment-based matching algorithm that allows a semantic frame to be partially matched through a statistical scoring scheme. Experiment results demonstrate that our approach can effectively detect readers' emotion by exploiting the syntactic structures and semantic associations in the context as well as outperforms currently well-known statistical text classification methods and the stat-of-the-art reader-emotion detection method.

雲端的盛行使得人們做任何事都要透過網路，但是總會有些有心人士使用一些惡意程式來創造攻擊或通過網絡連接竊取資料。為了防止這些網路惡意攻擊，我們必須不斷檢查網路流量資料，然而現在這個雲端時代，網路的資料是非常龐大且複雜，若要檢查所有網路資料不僅耗時而且非常沒有效率。 本研究使用TensorFlow與多個圖形處理器(Graphics Processing Unit, GPU)來實作類神經網路(Artificial Neural Networks, ANN)機制，用以分析網路流量資料，並得到一個可以判斷正常與異常網路流量的偵測規則，也設計一個實驗來驗證我們提出的類神經網路機制是否符合網路流向異常偵測的時效性和有效性。 在實驗過程中，我們發現使用更多的GPU可以減少訓練類神經網路的時間，並且在我們的實驗設計中使用三個GPU進行運算可以達到網路流量異常偵測的時效性。透過該方法得到的初步實驗結果，我們提出機制的結果優於使用反向傳播算法訓練類神經網路得到的結果。 / The prosperity of the cloud makes people do anything through the Internet, but there are people with bad intention to use some malicious programs to create attacks or steal information through the network connection. In order to prevent these cyber-attacks, we have to keep checking the network traffic information. However, in the current cloud environment, the network information is huge and complex that to check all the information is not only time-consuming but also inefficient. This study uses TensorFlow with multiple Graphic Processing Units (GPUs) to implement an Artificial Neural Networks (ANN) mechanism to analyze network traffic data and derive detection rules that can identify normal and malicious traffics, and we call it Network Traffic Anomaly Detection (NTAD). Experiments are also designed to verify the timeliness and effectiveness of the derived ANN mechanism. During the experiment, we found that using more GPUs can reduce training time, and using three GPUs to do the operation can meet the timeliness in NTAD. As a result of this method, the experiment result was better than ANN with back propagation mechanism.

摘要 本研究主要利用誤差偵測方法發掘數值高程模型中可能出現的高程誤差，藉以提升數值高程模型之高程品質。本研究採用三種誤差偵測方法即參數統計、水流方向矩陣、坡度與變化約制等，這三種方法過去是應用在航測資料測製之格網式數值高程模型，本研究嘗試推廣至空載光達製作的數值高程模型。 利用模擬DEM資料以驗證三種偵測方法之偵測能力。首先利用多項式函數擬合出各種地形，並假設該地形無誤差。再將人為誤差隨機加入模擬DEM資料；第二部份則將誤差偵測之方法應用至真實的數值高程模型資料，並配合檢核點高程測量檢驗之。根據誤差偵測結果，參數統計和坡度變化結果類似而且皆有過度偵測之缺點，可透過提高門檻值或高通濾波改善；水流方向矩陣比較不適合誤差偵測，但可透過窪地填平最佳化地形。 關鍵字：數值高程模型、誤差偵測、參數統計法、坡度與變化約制、水流方向矩陣。 / Abstract In this study, error detection methods were proposed to find possible elevation errors in digital elevation model (DEM), and to improve the quality of DEM. Three methods were employed to detect errors in the study, i.e. parametric statistical method, flow direction matrix, and constrained slope and change. These methods can deal with grid DEM from photogrammetric approach in the past, and now the methods are used to find errors in high resolution DEM from light detection and ranging (LIDAR). The simulated DEMs were used to approve the detection capability of the proposed methods. The fitted DEMs were first obtained by polynomial functions fit the different terrains and assuming these DEMs were free of errors. Then the artificial errors were added to fitted DEMs. The proposed methods were also applied to real DEM data got from LIDAR and field check works were run to insure the results. The results of parametric statistical method and constrained slope and change are similar, and all show the over-detection of errors. These results can be improved by using high threshold or high-pass filter. Flow direction matrix is not suitable for error detection in DEM, but can be applied to fill sink to optimize terrain for watershed analysis. Keyword: digital elevation model, error detection, parametric statistical method, constrained slope and change, flow direction matrix.

近年來穿戴相關發展越來越蓬勃，特別是在虛擬-實境的綜合藝術表 演中，例如: 電影「猩球崛起」。然而大部分虛擬實境的綜合內容是 基於腳本預先錄製好的，而且演員需要大量的練習，使表演能夠完美 演出。此外，如果我們想要在兩人的相對方向定位上有特殊效果，那 麼預先錄製的的方法是不合適的。解決這個問題的一個方法是，使用 高品質的相機偵測身體的姿勢或位置。但是精准度常會受限於光線或 是障礙物。 本篇論文中，我們提出一個即時相對方向定位方法，這方法使用無線 可穿戴式設備解決這個問題。我們結合BLE 所發送的Received Signal Strength Indicator (RSSI)與IMU 感測器資訊，來追蹤兩個表演者的相對方向定位的位置。但是RSSI 資料有波動與不穩定性、IMU 會引起 累積的誤差。我們發明了「可靠程度」的RSSI 量測概念，並且把這 概念運用在IMU 定期校正上。我們實驗的情況是，兩個人的舞蹈來 驗證準確性，結果是令人滿意的。我們還使用Unity 來實踐人體骨架， 以便與兩個舞者動作做比較。在未來，我們開發的方案可以用於藝術 表演，使內容更豐富，更具互動性。 / In recent years, wearable-related applications are flourishing, especially in virtual-real integrated art performance, such as “Rise of the Planet of the Apes”. However, most of the virtual-real integrated contents are pre-recorded based on the script, and the performer needs a lot of practice to make the integration perfect. Moreover, if we want to make special effect based on the relative directional positions of two performers, the pre-recorded approach is not suitable. One way to tackle this problem is to use the high-quality camera to detect the body posture or position.But the accuracy is usually limited in light intensity or obstacles. In this thesis, we propose a real-time relative directional positioning approach using wireless wearable devices to solve this problem. We use Received Signal Strength Indicator (RSSI) of BLE, combined with IMU sensors to track two performers’ relative directional positions. The RSSI fluctuates and the IMU causes accumulated errors. We invent the concept of “reliable level” of RSSI measures to periodically correct the IMU errors. We experiment the scenario of two-person dance to validate the accuracy, and the result is satisfactory. We also use Unity to real-time render the human skeleton for comparison with the two dancers’ motion.In the future, our developed scheme can be used in the art performance to make the content richer and more interactive.