National Chung Hsing University
National Chung Hsing University Institutional RepositoryNot a member yet
82070 research outputs found
Sort by
A Study about the Relationship of Stories in Children Picture Books and the Development of Local Traditional Industry in Taiwan
No full text本論文以台灣兒童圖畫書的地方傳統產業為研究主軸，近年來，台灣兒童圖畫書以在地產業為主題的出版數量逐漸增加，除了政府政策的推廣之下，民間出版社也相繼出版，替產業圖畫書的內容增添不少色彩。
本文依照政府政策、社區營造以及民間獨立出版三大面向之文本為主，以創作背景與理念做深入的了解，接著藉由細讀文本之故事融，來做統整分析，並找出其可以討論的現象與議題。
本論文共分為七章，第一章〈緒論〉詳述研究地方傳統產業圖畫書的動機與目的、前人研究成果及研究方法與步驟。第二章為〈產業故事之形成背景〉，以台灣解嚴後所實施的教育政策、社區總體營造之政策以及民間出版社所出版的風格類型，進行地方傳統產業圖畫書的出版由來分析。第三章〈地方產業敘事與圖像表現方式〉，以地方產業的敘事視角、情境與知識層面，來分析產業圖畫書的教育意涵與多元的表現方式。第四章〈產業故事與人的情感認同〉藉由柔性的情感層面，包含家庭親情與鄉土情感，了解地方產業的年代意義與精神。第五章〈以本土產業故事來行銷社區形象與產品價值〉，透過社區營造的部分，讓故事呈現出社區產品的價值，以及居民對於社區的熱愛與投入。第六章〈從產業故事看社會文化現象〉，藉由民俗信仰與社會環境的變遷兩部分，了解文本裡面產業的傳統習俗與社會大環境改變的種種面向，看見產業圖畫書內容的不同面貌。第七章〈結論〉，透過故事的傳播，能帶領兒童體會濃厚的在地情，並拉近與土地的距離，同時使得台灣夕陽工業與民俗文化喚起政府與民眾的重視。運用人文敘事之手法，並結合文化與設計，圖畫書成為一種推廣地方傳統產業的方式。This study is talking about t the relationship of Stories in children picture books and the development of local traditional industry in Taiwan. Recently, The number of publication with the theme of Taiwan children''s picture books gradually increased, in addition to the publication of government policies, nongovernmental publishing houses have also published. The situation has enriched the variety of the style of Taiwan children''s picture books.
To analysis the background and the inspiration of these children picture books, I take the government policy, community construction and private independent publication as my references. I divided the study into seven chapters. The first chapter is about the reason to write this study and the method to complete the study. The second chapter is about the background and the formation of industrial stories in these children’s picture books. The third chapter is about how to show the stories through letters or pictures in these children’s picture books. In chapter four “the relationship of industry story and human emotional identity”, I tried to analysis the meaning and affection between family and feelings of the hometown and the development of local industry. Chapter five is talking about how to marketing the image of communities and the product from communities through these children’s picture books. Chapter six is talking about the observation of social and cultural phenomenon through the industrial stories in these children’s picture books. And I made my conclusion of the study in chapter seven.第壹章 緒論………………………………………………………………………001
第一節 研究動機與目的……………………………………………………001
第二節 文獻探討……………………………………………………………002
第三節 研究範疇與方法取徑………………………………………………006
第貳章 「產業故事」之形成背景………………………………………………015
第一節 台灣鄉土教育與教材的強化………………………………………015
第二節 社區意識的發展與產官學的連結…………………………………018
第三節 生命教育結合產業故事的盛行……………………………………021
第參章 地方產業敘事與圖象表現方式…………………………………………023
第一節 展現產業面貌的不同視角…………………………………………023
第二節 展現產業樣態的不同情境…………………………………………033
第三節 展現產業知識的不同方式…………………………………………044
第肆章 產業故事與人的情感認同………………………………………………058
第一節 家庭親情的展現……………………………………………………059
第二節 鄉土情感的展現……………………………………………………071
第伍章 以本土產業故事來行銷社區形象及產品價值…………………………081
第一節 透過產業故事介紹社區形象特色…………………………………082
第二節 透過產業故事推廣本土優質產品…………………………………089
第陸章 從產業故事看社會文化現象……………………………………………100
第一節 民俗信仰的展現……………………………………………………100
第二節 社會環境的變遷……………………………………………………109
第柒章 結論………………………………………………………………………119
參考書目：…………………………………………………………………………121
附錄：………………………………………………………………………………12
A Study of the Philosophy about Wang Yang-ming and Literature during Ming Dynasty
No full text對於一個真實面對自我與追求道德學問的思想家，我們跟隨著他的腳步，探訪其早年的生命經驗，瞭解王陽明思想發展的歷程。王陽明思想發展歷程裡，最關鍵的轉折點——「龍場悟道」，疾病、死亡如影隨形地逼迫著此時的王陽明。對死亡的超越，在這場大徹大悟之中，發現心的無限潛能，也展現陽明心學的獨特之處。完成了一次思想的飛昇。至龍場而後確立了良知之學的核心理論。第三章開始是陽明主要理論的分析討論，目的是歸納出一個新的儒學系譜——「力度儒學系譜」，透過這一系譜的特點與思想家的血緣性，企圖找出影響中晚明文藝場域的種種元素。新的儒學圖像在明代形成，不同於明初朱子學的體系，一個以道德情感為根基的「力度儒學」系譜。關於「力度儒學系譜」的定義，採用「力度」二字表示對於「心」當中情感與理性的合一，不排除情感內容，使道德實踐本身更加貼近生活世界，道德也不是割裂人性特質後蒼白無力的外在規範，尋回人本身道德主體的力量。中晚明的文藝場域受到陽明心學的影響，行動者產生什麼樣的生存心態，又如何在這個場域當中展現自我，進而對權力場域展開什麼樣的抵抗，則是第四章所欲探討的。將焦點鎖定在受陽明心學影響較深，且在明代具有代表性的文藝場域行動者作為分析的對象，瞭解他們如何以一種狂士的人格於文藝場域，透過文學藝術的關注，追求主體性、將自我與審美結合為一種藝術化生活。The study is intended to uncover the development of Wang Yangming''s thoughts by following the footsteps of a thinker who faced his true self in his pursuit of moral knowledge, and exploring the experience of his early life. The most critical turning point in the development of Wang Yangming''s thoughts occurred at Longchang. It was recorded in the history as 'The Enlightenment at Longchang'. When that event occurred, Wang Yangming''s life was threatened by illnesses and death. It was during the process of the Enlightenment, Wang discovered the unlimited potential of 'the mind (Xin)' and transcended death. The breakthrough demonstrated the uniqueness of 'the philosophy of the mind (Xin Xue)'. 'The Enlightenment at Longchang' achieved an uplifting of his thoughts. After Wang''s arrival at Longchang, his core theory of 'Good Conscience' was established. Starting from the third chapter, the Yangming Theory was fully discussed and analyzed, and the purpose was to induced a new Confucian pedigree -- the pedigree of Dynamic Confucian.
It is expected that the various elements of the literary realm of the middle and late Ming dynasty could be identified through the examination of the characteristics of this particular pedigree and the lineage. New Confucian topography, formed in Ming dynasty, was different from the lineage of Zhu Xi''s. The Yangming lineage is a dynamic Confucian lineage rooted in moral compassion. With regard to the definition of the 'dynamic Confucian lineage', the 'dynamic' is used to express the combination of rationality and emotion in the mind without getting rid of the emotion factor so that the practice of morality is close to living realm. Morality should not be a set of pale and listless specification separated from human nature. Instead, humans should be the subject of morality. The literary realm of the middle and late Ming dynasty was affected by Wang''s philosophy of the mind. What was the mentality of the creators of literary works? That was discussed in Chapter Four. Topics also discussed in Chapter Four: How did they express themselves? What was the resistance they cast upon the authority? The focus of the study was on those literary creators who were affected relatively deeper by the Yangming thoughts and whose works were characterized by their representative roles in the realm of literature. The study tries to understand how they lived by combining the expression of themselves and the beauty of their creative works into their artistic lives.第一章 緒論 ................................................ ........................................................1
第一節 研究動機 ....................................... .......................................................1
第二節 文獻探討與研究成果回顧....... .............................................................3
一、心學與理學大方向的比較研究........ ........................................................3
二、陽明思想整體的研究 ............... ...............................................................3
三、陽明思想個別概念的研究..........................................................................5
四、陽明思想與其他哲人思想的比較研究......................................................6
五、陽明思想的影響研究或周邊研究 ............................................................7
第三節 研究目的與方法 ...................................................................................8
第二章 陽明心學的發展歷程...............................................................................11
第一節 童年與少年時期的經歷................................. .......................................11
一、童年的奇異傳說：瑞雲降生與五歲不言...................... .........................11
二、遠大的志向：天地間第一等事................................................................12
三、光芒掩詩文：賦金山詩、成立詩社、官署學書......................................12
四、鑽研兵法統馭：出遊居庸三關與夢謁伏波將軍廟..................................16
第二節 陽明思想的發展歷程............... .............................................................18
一、早期思想....................................................................................................19
(一)格竹致疾：理學的嘗試.............................. ............................................19
(二)陽明洞修道：佛、老的探險....................................................................26
(三)體悟孝弟之仁：儒家的歸向.......................... ........................................33
二、龍場悟道...................................................................................................38
(一)政治風暴：直幹歷冰雪，密葉留清風 ..................................................38
(二)謫居龍場：君子居之，何陋之有？.........................................................42
第三章 陽明心學的主要理論與力度儒學系譜...................................................51
第一節、陽明心學的主要理論...........................................................................51
一、形上形下的破除：「心即理」對程朱理學的反動.....................................51
二、格物致知與四句教...... ............................................................................57
三、「心外無物」與意向性理論.......................................................................60
(一)「物」的詮釋與心物關係........................................................................61
(二)「意」的作用與意向性理論................................ ...................................63
第二節、陽明後學之流派與力度儒學系譜............................. ..........................66
一、四句教的詮釋與陽明後學之流派............................. ..............................66
(一)浙中學派.............................. ....................................................................72
(二)江右學派......................................... .........................................................74
(三)泰州學派................................ ..................................................................75
二、良知之學與力度儒學系譜..........................................................................76
(一) 良知之學的思想源流.................. ............................................................76
1.性善的根源：孟子的良知良能.................................................................76
2.先立其大：陸象山的道德本心.................................................................78
(二) 力度儒學系譜.................................................... ......................................81
第四章 陽明心學對中晚明文藝思潮的啟蒙與影響.............................................87
第一節 狂者的道路：批判精神與藝術化主體.................................................87
一、批判精神：李贄「童心說」與知識的解構............................................87
(一)狂者精神與力度儒學.................................... ..........................................87
(二)文藝場域中的抵抗精神............................................... ...........................92
二、審美主體與藝術化生活................... .......................................................100
(一)公安派文學中的審美主體.. ..................................................................101
(二)閒賞美學與藝術化生活的追求......... ...................................................104
第二節 自然情性的尋求與生活世界的呈現.................... ............................107
一、至情之說體現的象徵世界：湯顯祖與《牡丹亭》..............................108
二、通俗文學對生活世界之呈現：馮夢龍與「情教說」..........................113
第五章 結論.........................................................................................................119
參考文獻.................................................................................................................123
圖目次
圖2-1：(明)王守仁行書〈龍江留別詩卷〉..............................................................15
圖2-2：(明)王守仁行書〈與鄭邦端尺牘〉..............................................................16
圖2-2：(明)王守仁行書〈銅陵觀鐵船歌卷〉...........................................................16
圖4-1：陽明心學師承關係與力度儒學系譜圖.......................................................91
圖4-2：布爾迪厄「場域結構分析圖」....................................................................9
Application and Study of imbalanced datasets base on Top-N Reverse k-Nearest Neighbor (TRkNN) coupled with Synthetic Minority Over-Sampling Technique (SMOTE)
No full text不平衡分類是指數據集具有不均勻的類別分佈。若不考慮數據集的不平衡問題，大多數分類方法對於多數類的預測有較高的準確率，而少數類的準確率則明顯較低。本研究第一項工作是提出一個有效的演算法，此演算法結合反向排序K鄰近法(TRkNN)和合成少數類樣本的增量技術(SMOTE)，以克服UCI資料庫中不平衡數據集的問題。為了研究此演算法，本研究也將其應用於不同的分類方法，如邏輯回歸、C4.5、SVM和BPNN。此外，還採用不同的距離度量來分類相同的UCI數據集。經驗結果表明，歐幾里德距離和曼哈頓距離不僅具有更高的準確率，而且還有比切比雪夫距離和餘弦距離更快的計算效率。因此，基於TRkNN和SMOTE的演算法可以廣泛用於處理不平衡數據集，如何選擇合適的距離度量可以作為未來研究的參考。
對癌症預測的研究已經應用多種機器學習演算法，如類神經網絡，基因演算法和粒子群演算法，以找出分類疾病或癌症的關鍵屬性或傳統的統計預測模型，有效地區別不同類型的癌症，從而建立可以提早發現和治療的預測模型。其中以現有患者的資料作為訓練集來建立模型以預測新病患樣本的分類準確度。這個問題在資料探勘領域引起了相當大的關注，學者們提出了各種方法（例如，隨機抽樣和特徵選取）來解決類別不平衡並實現重新平衡的類別分佈，從而提高分類器的有效性。雖然重新採樣方法可以快速處理不平衡樣本的問題，但它們更重視多數類中的數據，忽略少數類中潛在的重要數據，從而限制分類的有效性。根據在不平衡醫學數據集中發現的模式，本研究第二項工作是使用合成少數類樣本的增量技術來改善不平衡數據集的問題。此外，這項研究還使用三個UCI醫療數據集來比較基於機器學習、軟計算和仿生計算之各種方法的重新採樣性能。The imbalanced classification means the dataset has an unequal class distribution among its population. For a given dataset without considering the imbalanced issue, most classification methods often predict the high accuracy for the majority class, but significantly low accuracy for the minority class. The first task in this dissertation is to provide an efficient algorithm, Top-N Reverse k-Nearest Neighbor (TRkNN), coupled with Synthetic Minority Over-Sampling TEchnique (SMOTE) to overcome this issue for several imbalanced datasets from famous UCI datasets. To investigate the proposed algorithm, it was applied into different classified methods, such as Logistic regression, C4.5, SVM, and BPNN. In addition, this research also adopted different distance metrics to classify the same UCI datasets. The empirical results illustrated that the Euclidean distance and Manhattan distance not only perform higher percentage of accuracy rate, but also show greater computational efficiency than the Chebyshev distance and Cosine distance. Therefore, the TRkNN and SMOTE based algorithm could be widely used to handle the imbalanced datasets and how to choose the suitable distance metrics can be as the reference for the future researches.
Research into cancer prediction has applied various machine learning algorithms, such as neural networks, genetic algorithms, and particle swarm optimization, to find the key to classifying illness or cancer properties or to adapt traditional statistical prediction models to effectively differentiate between different types of cancers, and thus build prediction models that can allow for early detection and treatment. Training data from existing patients is used to establish models to predict the classification accuracy of new patient samples. This issue has attracted considerable attention in the field of data mining, and scholars have proposed various methods (e.g., random sampling and feature selection) to address category imbalances and achieve a re-balanced class distribution, thus improving the effectiveness of classifiers with limited data. Although resampling methods can quickly deal with the problem of unbalanced samples, they give more importance to the data in the majority class, and neglect potentially important data in the minority class, thus limiting the effectiveness of classification. Based on patterns discovered in imbalanced medical data sets, the second task in this dissertation is to use the synthetic minority oversampling technique to improve imbalanced data set issues. In addition, this research also compares the resampling performance of various methods based on machine learning, soft-computing, and bio-inspired computing, using three UCI medical data sets.Contents
致 謝 ii
中文摘要 iii
Abstract iv
Contents vi
List of Figures viii
List of Tables ix
Chapter 1 Introduction 1
Chapter 2. Literatures Review 4
2.1 Sampling Techniques 4
2.2 Synthetic Minority Oversampling Technique (SMOTE) 5
2.3 Machine Learning 6
Chapter 3. Data Mining for Bioinformatics: Design with Oversampling and Performance Evaluation 8
3.1. Imbalanced Class 8
3.2. Machine Learning 10
3.3. DataBase 11
3.4. Methods 13
3.4.1. UCI Data Set Collection Stage 13
3.4.2. Data Preprocessing Stage 13
3.4.3. Prediction Model Implementation Stage 15
3.4.4. Performance Evaluation Stage 15
Chapter 4. Distance Metric Based Over-Sampling Method for Bioinformatics and Performance Evaluation 16
4.1. Materials 16
4.2. Top-N Reverse k-Nearest Neighbor (TRkNN) Algorithm 17
4.3. Distance Metrics 19
Chapter 5. Experimental Results 22
5.1. Performances of Data Mining Based Design for Bioinformatics with Oversampling 22
5.1.1. Experimental Design and Parameter Setting 22
5.1.2. Performance Evaluation 22
5.2. Performances of Distance Metric Based Over-Sampling Method for Bioinformatics 25
5.2.1. Experimental Design and Parameter Setting 25
5.2.2. Performance Evaluation 25
Chapter 6 Conclusions 31
References 3
The Assessment Function of Diabetes Mellitus Risk
No full text糖尿病為代謝異常的疾病，隨著生活水準逐漸升高亦跟著逐漸增加的趨勢，糖尿病的症狀有：頻尿、口渴、飢餓感、容易疲勞等症狀，最初容易被患者忽略而尚未就醫，即使就醫也極有可能因為患者尚未注意某些症狀而未告知醫療人員，不但會造成醫療人員對於診斷上的誤判而造成患者錯失就醫的良好時機，造成許多不可逆的生理影響，使得患者的生活品質大幅下降。糖尿病必需檢驗尿糖與血糖才能確認是否罹患糖尿病，況且若糖尿病沒及時處理而導致病情惡化恐怕會因此產生許多的併發症，如：心血管疾病、視網膜病變、腎病變等許多嚴重的疾病。
本篇論文針對糖尿病建立估計模組，收集多位參與者的數個mRNA特徵值，使用Fisher Ratio求得得分函數再運用多元迴歸分析應用來建立糖尿病評估模型，醫療人員可藉由模型運算出來的結果作為參考，並給予患者適當地檢驗來確認患者是否已得了糖尿病，如此的參考資料不但增進了醫療人員的效率，亦能減少因誤判而錯失時機造成糖尿病惡化的結果。Diabetes mellitus is metabolic abnormalities of disease. Diabetes mellitus gradual increment trend with advance of the living standard at different succession stages. There has frequent urination、thirsty all the time、sense of hunger、tiredness that the symptoms by diabetes mellitus what the symptoms of the patient ignore at the first cause not used in treating. Medical personnel didn''t defendant know the patient has diabetes mellitus at high risk cause the patient not aware of those symptoms immediately that medical personnel was erroneously diagnosed and the treatment delayed. Delay medical treatment of diabetes mellitus result in irreversible damage to physiological that make the patient drop quality of life down. A special medical inspection of the diabetes mellitus with urine sugar test and blood sugar test to make sure the patient has diabetes mellitus. The patient has deteriorated with no treatment as soon as possible make many complications set in, got serious illnesses like cardiovascular disease、diabetic retinopathy、renal disease.
This article configuration an evaluation of the diabetes mellitus model there are collected many participants datum includes several mRNA feature value. Apply Fisher Ratio got the score function and multiple regression analysis building the evaluation of diabetes mellitus model. Medical personnel by mathematical operation of the evaluation model of reference material for arrange the appropriate inspection items make sure the patient has diabetes mellitus that speed up efficiency for medical personnel and low misdiagnosed down of diabetes mellitus patient.摘要 i
Abstract ii
目錄 iv
圖目錄 vi
表目錄 x
第一章 緒論 1
1.1 研究背景(Research Background) 1
1.2 研究動機與目的(Research Motives and Purposes) 5
1.3 論文架構(Paper Structure) 6
第二章 相關文獻 7
2.1 相關研究(Related Research) 7
2.1.1 生物統計(Biostatistics) 8
2.1.2 特徵值(Eigenvalue) 9
2.2 相關技術(Related Technology) 10
2.2.1 Fisher Ratio Classification Method 11
2.2.2 累積分佈函數(Cumulative Distribution Function) 12
2.2.3 常態分佈(Normal Distribution) 13
2.2.4標準常態分佈(Standard Normal Distribution) 15
2.2.5 多元迴歸分析(Multiple Regression Analysis) 16
第三章 實驗方法 21
3.1 程式流程(Program Flow) 21
3.2 資料設定(Data Set) 22
3.3 標準化(Standardization) 25
3.4 得分函數(Score Function) 36
3.5 資料分佈(Data Distribution) 38
第四章 實驗結果 40
4.1 分類結果(Classification Results) 40
4.2 資料轉換(Data Transformation) 57
4.3 陽性函數及陰性函數(True Positive and False Negative Function) 59
4.4 評估結果(Evaluation Result) 60
第五章 結論與未來展望 63
參考文獻 65
圖目錄
圖1-1 男女失業率 3
圖1-2 全台22縣市歷年糖尿病比例[4] 4
圖2-1 蒙娜麗莎映像與原圖 10
圖2-2 常態分佈曲線圖 15
圖2-3 Z的常態分佈涵蓋機率 16
圖3-1糖尿病評估模型流程圖 21
圖3-2 糖尿病取樣男女比例圖 22
圖3-3 即時聚合酶連鎖反應 23
圖3-4 聚合酶連鎖反應 24
圖3-5 DNA與RNA 25
圖3-6 第一特徵年齡與特徵值分佈圖 26
圖3-7 第二特徵年齡與特徵值分佈圖 27
圖3-8 第三特徵年齡與特徵值分佈圖 27
圖3-9 第四特徵年齡與特徵值分佈圖 28
圖3-10 第五特徵年齡與特徵值分佈圖 28
圖3-11 第六特徵年齡與特徵值分佈圖 29
圖3-12 第七特徵年齡與特徵值分佈圖 29
圖3-13 第八特徵年齡與特徵值分佈圖 30
圖3-14 第九特徵年齡與特徵值分佈圖 30
圖3-15 第十特徵年齡與特徵值分佈圖 31
圖3-16 第十一特徵年齡與特徵值分佈圖 31
圖3-17 第十二特徵年齡與特徵值分佈圖 32
圖3-18 第十三特徵年齡與特徵值分佈圖 32
圖3-19 第十四特徵年齡與特徵值分佈圖 33
圖3-20 第十五特徵年齡與特徵值分佈圖 33
圖3-21 第十六特徵年齡與特徵值分佈圖 34
圖3-22 第十七特徵年齡與特徵值分佈圖 34
圖3-23 第十八特徵年齡與特徵值分佈圖 35
圖3-24 第十九特徵年齡與特徵值分佈圖 35
圖3-25 第二十特徵年齡與特徵值分佈圖 36
圖3-26 年齡與分數分佈圖 39
圖4-1 第1∼4特徵未分類箱型圖 41
圖4-2 第5∼8特徵未分類箱型圖 41
圖4-3 第9∼12特徵未分類箱型圖 42
圖4-4 第13∼16特徵未分類箱型圖 42
圖4-5 第17∼20特徵未分類箱型圖 43
圖4-6 第一特徵分群箱型圖 44
圖4-7 第二特徵分群箱型圖 44
圖4-8 第三特徵分群箱型圖 45
圖4-9 第四特徵分群箱型圖 45
圖4-10 第五特徵分群箱型圖 46
圖4-11 第六特徵分群箱型圖 46
圖4-12 第七特徵分群箱型圖 47
圖4-13 第八特徵分群箱型圖 47
圖4-14 第九特徵分群箱型圖 48
圖4-15 第十特徵分群箱型圖 48
圖4-16 第十一特徵分群箱型圖 49
圖4-17 第十二特徵分群箱型圖 49
圖4-18 第十三特徵分群箱型圖 50
圖4-19 第十四特徵分群箱型圖 50
圖4-20 第十五特徵分群箱型圖 51
圖4-21 第十六特徵分群箱型圖 51
圖4-22 第十七特徵分群箱型圖 52
圖4-23 第十八特徵分群箱型圖 52
圖4-24 第十九特徵分群箱型圖 53
圖4-25 第二十特徵分群箱型圖 53
圖4-26 糖尿病患者分數直方圖 57
圖4-27 正常受試者分數直方圖 58
圖4-26 常態分佈圖 59
圖4-27 陽性函數ΦDMx與陰性函數ΦNOx曲線圖 60
圖4-28 F-Factor曲線圖 62
表目錄
表4-1 各個特徵值未剔除離群值之平均值與標準差 54
表4-2各個特徵值剔除離群值之平均值與標準差 5
A Method of Non-Contact Proximity Detection for Table Saws
No full text近代社會的快速進步濫觴於工業革命，到了二十世紀逐漸出現大量電動機具的應用，電動機具的強勁動力能幫助作業更有效率地進行，卻也使得操作過程中伴隨意外發生的危險。以圓鋸機為例，大量運用在木板加工過程，被利用於對木板進行直線切割等裁切作業，但是也容易發生身體部位，例如手部，會與運作中的圓鋸片刀刃過於接近而造成意外發生。因此若能夠及時得知此狀況的發生，即可以早先一步於接觸意外發生前發出警示訊號，讓操作者有更多反應時間，以盡可能降低意外損害或是避免發生。
本篇論文主軸為對平台式圓鋸機建立出接近偵測系統。利用將非接觸式感測元件設置在平台下方，當有物體接近時則進行數據蒐集，並建立出一套基於機率類神經網路(Probabilistic Neural Network，PNN)的分析系統以對擷取的數據進行分類判斷。以包含來自不同人的綜合資料集進行實驗，依序對單手及雙手的情況進行檢測，在各自的實驗中，分別能到達95%與89%的準確率。進一步本論文提出針對單人數據時的進階PNN學習模型，以強化相同情境下之辨識準確率，在此模型下辨識率能提升至97%。The rapid progress of modern society begin in Industrial Revolution. After into the twentieth century, the large number of electric applications gradual emerged. Electric power tools to help the operation more efficiently, but it also makes the operation accompanied by the risk of accidents. In the case of table saws, a large number of table saws are applied in the wood processing aspect, which is used for wood splitting such as cutting straight lines. However, it is often occur body, such as hands or fingers, is so close to the running saw blade that cause accidents. Therefore, if it can inform the occurrence as soon as possible, that can issue the warning in time. In doing so, operators can have more reaction time avoiding accidents occurred.
The backbone of this paper is establish the proximity detection for table saws. By setting the non-contact sensors, which collecting data when the object close to the saw. Simultaneously, establish an analysis system which is based on Probabilistic Neural Network (PNN) to determine the classification of collected data. Experimenting with a comprehensive data set from different people. Testing the situation of single hand and two hands sequentially. In each experiment, the accuracy can reach 95% and 89% individually. This paper further proposes a PNN learning-fixed model when the analysis system is used by single person. The accuracy can reach 97% in the case.誌謝辭 i
中文摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 viii
一、 緒論 1
1.1. 研究背景 1
1.2. 動機與目的 2
1.3. 論文架構 2
二、 文獻探討與相關研究 3
2.1. 平台式圓鋸機介紹 3
2.1.1. 平台式圓鋸機防護機構設置 3
2.1.2. 接觸災害情況描述 5
2.1.3. 平台式圓鋸機接觸式偵測 6
2.1.4. 基於接觸式觸碰偵測的切入距離研究 7
2.2. 感測元件的架構與原理 8
2.3. 機率類神經網路 10
2.3.1. 貝氏決策(Bayes’Rule)運用於PNN 11
2.3.2. Parzen視窗法 11
2.3.3. PNN的4層架構說明 13
三、 平台式圓鋸機非接觸式接近偵測方法 15
3.1. 待紀錄特徵選擇 15
3.2. 手勢定義 18
3.3. 特徵分析流程 22
3.3.1. 非接觸式接近偵測主要流程 22
3.3.2. 單次接近偵測過程 23
3.3.3. 資料前處理 23
3.3.4. 以基礎PNN網路進行分類最佳化 26
3.3.5. 進階PNN學習模型 27
四、 實驗結果 31
4.1. 實驗環境設置 31
4.1.1. 實驗儀器與工具 31
4.1.2. 感測器擺放距離 33
4.1.3. 資料收集流程 34
4.2. 單次PNN自我驗證 34
4.3. 迭代測試比較Spread值 35
4.3.1. 單手迭代測試 36
4.3.2. 單/雙手混合資料集迭代測試 41
4.4. 進階PNN學習模型結果 46
五、 結論與未來展望 53
5.1. 結論 53
5.2. 未來展望 53
參考文獻 5
A Study on The IR-Image Based Driver Monitoring System Using Facial Landmarks
No full text近年來，昏昏欲睡(Drowsiness)以及分心(Distraction)皆為造成車禍屢屢發生的主要因素，其中，造成注意力不集中的原因可能為：在車內飲食、與其他人講話、使用3C產品、疲勞駕駛等等，因此發展一個有效的駕駛員注意力監測系統已成為預防車禍發生的重要議題。我們可以藉由分析駕駛的眼睛以及頭部動作來衡量駕駛員注意力不集中的程度，並適時地提供警訊，即可預防駕駛員因不專心或疲勞而造成嚴重的交通事故。
隨著日新月異的科技，雖然已發展出許多駕駛員注意力監測系統(Driver-inattention monitoring systems, DIMSs)，但大部分系統並未能同時監測疲勞以及分心的狀態，因此本論文提出一個駕駛員監測系統，目標即為將兩種狀態皆辨別出來。首先，本研究採用樹梅派的NoIR相機作為取像的儀器，使得系統在白天及晚上皆能運作。使用著名的Viola-Jones[10]演算法偵測影像中的人臉的位置，並使用臉部標記法標記出人臉的五官，接著使用主成分分析(PCA)及線性判別分析(LDA)萃取眼睛特徵並結合支援向量機(SVM)分辨出眼睛、非眼睛、開眼以及閉眼做為本研究監測疲勞的演算法；使用人臉標記法標記的座標位置來當作判斷分心與否的依據。
實驗環境皆為在車上駕駛，經由實驗結果顯示，本系統在偵測眼睛的部份不論是白天或晚上正確率皆達90%以上，在偵測眼睛狀態的部分上也有85%以上的正確率。In recent years, there are many traffic accidents caused by drowsiness and distraction. Eating, drinking, speaking someone on the phone and typing a text message etc. while driving, those are the mostly reason for the inattention. Therefore, developed a robust driver monitoring systems is the big issue to prevent car accidents. This kind of system will detect driver drowsiness or distraction level, which can be analyzed by driver’s eyes and head movement. Our system will alarm the driver if the drowsiness or distraction level is high. Therefore, we propose a driver monitoring system using facial landmarks based on IR-image. First, we use Raspberry Pi No Infrared (NoIR) camera to get the video which can work daytime and nighttime. Then we use the Viola-Jones detector[6] to detect the face. Second, we use facial landmarks method to locate the facial features. In distraction detection, we use one coordinate of the face to classify whether the driver is distracting or not. If the driver is distracting then the drowsiness detection is not performed. In drowsiness detection algorithm, we combine PCA and LDA to extract eye features then classified by SVM to determine whether the eye region is the eye or not. Next, using PCA and LDA combine with the area of the eye, and also classified the eye status (opened or closed) by SVM. Experimental results demonstrate that our eye detection can achieve 90% accuracy in both daytime and nighttime, and in the eye status detection with a success rate of 85%.第一章 緒論 1
1.1 研究背景與動機 1
1.2 論文架構 4
第二章 文獻探討 5
2.1基於駕駛員駕駛車輛的行為(Driving behavior) 5
2.2基於駕駛員本身的行為(Driver behavior) 6
2.2.1基於生理特徵法(Physiological feature-based) 6
2.2.2基於視覺特徵法(Visual feature-based) 7
第三章 研究方法 11
3.1圖像採集方法(Image acquisition method) 12
3.2 臉部偵測(Face detection) 14
3.3 眼睛偵測(Eye detection) 16
3.3.1 臉部標記法(Facial landmarks) 17
3.3.2 樣板比對(Template matching) 20
3.3.3 斑點檢測(Blob detection) 21
3.3.4 眼經驗證機制(Eye validation) 22
3.4 眼睛狀態偵測(Eye state detection) 26
3.5 分心偵測(Distraction detection) 28
3.6 疲勞偵測(Drowsiness detection) 30
第四章 實驗結果與討論 32
4.1 眼睛偵測實驗結果比較 32
4.2 眼睛狀態實驗結果比較 34
4.3 疲勞及分心實驗結果 36
第五章 結論與未來展望 39
參考文獻 4
An Efficient Flow-aggregation Scheme Based on FTRS Algorithm in SDN
No full text軟體定義網路（SDN）近幾年來發展迅速，也有許多關於SDN性能上的研究。不同於傳統的網路架構，SDN主要的概念是將網路程式化，為了使網路管理更為靈活，將交換器中的控制層分離出來。而我們主要關注於OpenFlow 交換器上流量表的大小，目前交換器的流量表是使用Ternary Content Addressable Memory (TCAM)，雖然TCAM查找速度快，但也有價格昂貴、功耗大的缺點，這些限制導致流量表的大小受到限制，因此可能發生流量表滿溢的問題，為了解決這個問題，我們會將流量表中的規則利用規則的特徵進行壓縮聚合，進而減少流量表的使用空間。
我們在進行流量聚合前，會根據欄位建立二元樹，每個節點會儲存規則。為了減少流量聚合的時間，會在每個節點建立一種索引，可知道每個規則可壓縮的範圍，以減少流量聚合的時間。為了改進流量聚合的壓縮率，會將流量表的規則依據行為作分類，再由每個分類中規則的數量多寡依序插入二元樹，透過規則的還原與數量大的規則優先進行流量聚合，讓規則的數量減少到最小。並且利用每個規則經過節點時會可知道聚合範圍的索引進行流量聚合，提升流量聚合過程的效率，減少尋找擁有同樣指令規則的時間成本，我們將改良後的方法簡稱為EFAS (Efficient Flow Aggregation Scheme)，為更加有效率的流量聚合方法。本篇論文在Linux平台上，使用Floodlight作為控制層中的控制器，並使用Mininet模擬OpenFlow Switch以及網路拓樸環境，實驗中我們使用了mesh、tree和fat tree三種拓樸，透過實驗結果得知，我們提出的索引以及規則分類方法的平均壓縮率較FTRS好約15%，而平均流量聚合時間較FTRS快了約2.0毫秒。The Software Defined Network（SDN）has been developed rapidly and there are numerous studied on the performance of SDN in recent years. Different from the traditional networks, SDN, which is the programmable networks, has more flexible network management which separate the controller plane away from the network switch. Our thesis focuses on the size of flow table in OpenFlow switch which is implemented by Ternary Content Addressable Memory (TCAM). Although TCAM has the advantage of fast forwarding, it is expensive and power hungry. These shortages limit the size of flow table and may cause the flow table overflow. To solve the problem, we will find the features of flow rules, and aggregate the rules according to the same features to reduce the size usage of flow table.
In this thesis, Flow Table Reduction Scheme (FTRS) is the subject matter we study and improve. Before aggregating the flow rules, we generate the binary tree and build range index on every node to identify the aggregation range, and the index can shorten the execution time of flow aggregation. To improve the compression ratio of flow aggregation, we classify the flow rules by instructions in each classification before inserting into the binary tree, and restore and classify the rules to decrease the number of flow rules to minimum size. The Scheme we proposed named Efficient Flow Aggregation Scheme, is the more efficient flow-aggregation method. With Linux operating system, we use Floodlight as the controller and use Mininet to simulate the OpenFlow Switch and network topology. Furthermore, by using mesh, tree and fat tree topologies in our simulation and it turns out that the average compression ratio of our method is relatively 15% better than that of the FTRS, and the average reduction of 2.0ms of the flow aggregation time.摘要 i
Abstract ii
內容 iii
表目錄 v
圖目錄 vi
第一章 緒論 1
1.1軟體定義網路(SDN)簡介 1
1.2研究動機和目的 3
1.3 論文架構 3
第二章 相關文獻探討 4
2.1 傳統網路IP路由表的流量聚合演算法 4
2.2 TCAM中封包分類表的流量聚合演算法 4
2.3 OpenFlow 交換器流量表的流量聚合演算法 4
2.4 FTRS研究 5
2.4.1 FTRS的資料結構 6
2.4.2 FTRS 流量聚合演算法 6
第三章 EFAS - IP位址流量聚合方法 8
3.1核心概念 8
3.2 EFAS的說明與舉例 9
3.2.1 規則的分類與排序 11
3.2.2 建立二元樹 15
3.2.3 範圍索引的使用與流量聚合的條件 15
第四章 模擬結果與分析 22
4.1 SDN模擬環境介紹 22
4.1 模擬結果與分析 24
第五章 結論 29
參考文獻 3
Surveillance System Design for Vehicle Tracking and VLSI Architecture Design of Feature Detection in Scale Invariant Feature Transform
No full text自動化影像應用系統在高解析影像串流愈發普及的現今生活中，佔有相當重要的輔助地位，小到家庭社區大到空拍航拍只要後台的電腦能夠支援該應用的運算複雜度，皆可實現系統的生活化與微型化。在諸多運用中追蹤系統通常是支撐複雜運用的核心之一，同時也是運算複雜度與準確度最難調和的區塊，越是要能達到高精確且高通用性，所需搭配的演算法也越加複雜多樣。
本架構為了提高影像的多物件追蹤精確率，導入高運算複雜度的局部影像特徵轉換SIFT，在實作中將解決如何篩選並轉換比對後特徵間的幾何關係，並利用校正之幾何區間定義出新畫面中的比對物件，利用不斷重複不同物件間的上述流程達到精確定位與持續追蹤的效果。
實驗中發現，雖然SIFT之特徵比對具128維的特徵向量作為描述，但仍會在與預測區間特徵比對中產生數組的錯誤比對，如何利用其幾何關係篩選出更為可靠的比對組做幾何轉換式本架構中相當重要的部分，在經過一系列的嘗試後以向量相關性作為判斷依據，大幅減少錯誤比對造成的轉換錯誤。
本系統經過軟體模擬後在功能性上達到預設需求，但運算量過於龐大造成無法即時運算，為有效加速以達到實際需求，將SIFT中的特徵檢測功能剝離出做硬體化設計，將其高度平行化運算特性加速尺度空間的建立與梯度圖的產生，並在最後的達成其工作預設需求。Nowadays, automatic visual system with high resolution video stream application is much more common in our life. With huge progress of computer and mobile system, we can use this powerful tool to help us conquer the massive computation of visual analysis and their related applications. Amount the visual system, objects tracking is almost the most basic but complicated subject, the user always wants to find the perfect balance between computation and precision, with more complex application, we used to find more and more new algorithms to solve unexpected problems.
In this architecture, in order to increase the accuracy of multi-object tracking, I use the scale invariant feature transform to establish the ID of each registered objects. After matching, all the features in database with searching area, the major problem is to find the robust pairs of those matching key points. With this key points, I can find the precise transform matrix to locate the update set of key points in searching area. Repeat all this rule to find each relocate objects in the new input frame.
Because of massive computation, I have to speed up a part of my design to catch up the real time implementation requirement. So I decide to build a hardware version of SIFT feature detection to replace the software one, take advantage of high parallelism of the algorithm of detection itself, the hardware can really reduce much of computation to speed up my original architecture.第1章 引言 1
1.1. 常見的影像來源與影像特性 1
1.2. 常見的自動化影像系統 2
1.3. 發展經歷 3
1.4. 舊版經驗總結 3
第2章 物件追蹤相關演算法介紹與探討 4
2.1. 影像追蹤與記算機視覺概述 4
2.2. Moving Objects Detection 5
2.3. Motion Estimation 6
2.4. 更具辨識性的物件資訊 7
第3章 基於SIFT特徵轉換之影像追蹤演算法 8
3.1. 多物件追蹤架構概述 8
3.2. 本架構之最初設計思路 9
3.3. 物件追蹤演算法流程圖 11
3.4. 特徵檢測演算法 14
3.4.1. 特徵描述特性概述 14
3.4.2. 常見之特徵演算法 15
3.5. 利用SIFT之特徵描述取代亮度特徵 16
3.5.1. SIFT特徵轉換演算法 16
3.5.2. 特徵比對 21
3.5.3. 對於追蹤物的描述 23
3.5.4. 快速比對與初步篩選 24
3.6. 物體中心定為與幾何校正 25
3.6.1. 第一版利用篩選特徵比對進行定位與校正 25
3.6.2. 改良版利用Matching Points計算兩平面的Homography Matrix 26
3.7. 快速矩形區間判定之更新演算法 29
3.8. 軟體演算法架構實驗規格 30
第4章 硬體架構設計與實作 39
4.1. 硬體規格 39
4.2. 硬體架構圖 40
4.2.1. 特徵偵測完整硬體流程圖 40
4.2.2. Frame Buffer 41
4.2.3. Convolution Architecture 45
4.3. Synthesis 合成 47
第5章 結論 51
參考文獻 5
Design of A Wireless Vehicle Charging System Adapted to Mutual Inductance Variation
No full text進十年來，世界各國的學者在無線充電的研究及技術應用上已發展出不錯的成果及相關應用，而其中應用無線充電於電動車輛的研究更是在工業及商業的發展中佔有一席之地。對於電動車輛充電技術，一大研究趨勢為應用感應式電能傳輸於電動車方面的研究。
目前在電動車無線傳輸領域內有兩項較為新穎的技術，分別是雙向無線電力傳輸及雙向數據傳輸。雙向無線電力傳輸允許車輛充電並供電給其他系統; 雙向數據傳輸給予充電站與電動車間彼此溝通的能力而不用任何附加的通信設備。兩項技術皆對於傳輸端與接收端間線圈相對位置敏感，然而傳輸線圈和接收器線圈之間的未對準會導致輸出功率下降，這種線圈不對稱的問題將嚴重影響電力傳輸效率、功率。
本論文於考慮線圈未對準問題下設計具有適應互感變動之雙向無線電力傳輸與雙向數據傳輸系統，並採用模糊規則來控制系統當線圈有未對準的情況。本論文採用SAE TIR 2954所推薦的頻率，傳輸頻率控制在85 KHz左右(81.3K-90KHz).In the past ten years, scholars from all over the world have developed results and related applications in the research and application of wireless charging, applications of wireless charging in the electric vehicles plays an important role in industrial and commercial development. In the electric vehicle charging technology, a major research trend is to adopt loosely coupled inductive power transfer in the electric vehicles.
There are two novel techniques in wireless power transfer region, bidirectional wireless power transfer and bidirectional data transmission. The capability of bidirectional wireless power transfer allows vehicles to be charged and provide charge to other systems; the capability of bidirectional data transmission allows WPT system charging without any additional communication device. Both of them are sensitive to precisely coil align. Due to misalignment between the transmission coil and the receiver coil, the output power might result in a drop. This misalignment problem will seriously affect the power transmission efficiency, communication quality.
This thesis analyzes and designs the coil misalignment problem and using fuzzy rule to solved coil misalignment problem under the system combining bidirectional wireless power transfer with the bidirectional data transmission. The system operates the switch frequency around 85 KHz, as recommended by SAE TIR 2954誌謝 i
摘要 ii
Abstract iii
Contents iv
List of Figures vi
List of Tables x
Chapter 1. Introduction 1
1.1 Survey of Related Research 2
1.2 Motivation and Objectives 3
1.3 Contributions 4
1.4 Brief Overview 4
Chapter 2. Wireless Power Transfer and Data transmission 5
2.1 Wireless Power Transfer 5
2.1.1 Far-Field Wireless Power Transfer 7
2.1.2 Near-Field Wireless Power Transfer 9
2.2 Bidirectional Power Transfer 14
2.3 Bidirectional Data Transfer 15
Chapter 3. Frequency-Spacing Modeling 18
3.1 Introduction 18
3.2 Coil Position Model 19
3.3 Equivalent LCIPT Circuit 22
Chapter 4. System Design 27
Chapter 5. Hardware Implement & Software Design 37
5.1 Introduction 37
5.2 System Hardware 38
5.2.1 Power Transfer Circuit 39
5.2.2 Driver Circuit 42
5.2.3 Measurement Circuit & Decoder Circuit 46
5.3 Control Strategy 48
Chapter 6. Experimental Results 54
Chapter 7. Conclusions and Future work 91
References 9
Automatic Water Meter and License Plate Recognition System using Extreme Learning Machine
No full text本論文是使用極限學習機(Extreme Learning Machines)進行水錶和車牌辨識，也使用了樣板比對、放射狀基底函數(Radial Basis Function)、支持向量機 (Support Vector Machine)做比較，傳統常使用的方法為樣板比對法，需要使用樣板和輸入影像一一比對，比對出最接近的得到結果，極限學習機(ELM)為神經網路只需要提出影像特徵，將影像特徵輸入神經網路中就能直接得到輸出結果，就能有效的提高運算速度，在車牌辨識中67個樣板(新式和舊式車牌)，進行40張車牌影像辨識，在使用樣板比對法運算時間大約為3秒，使用極限學習機(ELM)大約為0.2秒，在辨識速度上是明顯優於其他方法，在車牌辨識中使用(EI-ELM)辨識率可達87.5%，樣板比對法則是70%。This thesis paper uses Extreme Learning Machines for water meter and license plate recognition, and also uses template matching, Radial Basis Function and Support Vector Machine to compare with Extreme Learning Machines. The most traditional and commonly method is the template matching method, which needs to compare use the template and the input image to compare one by one, and the closest one is result is obtained.
The ELM neural network only needs to present the image features. and Put the image features are input to into the neural network then it can get the output result directly, In the output can be directly obtained, and also improve the speed of operation we can effectively. improve the speed of operation in The license plate identification recognition 67 models (new and old license plate), 40 license plate image recognition, the use of template comparison method of operation time is about 3 seconds, The use of ELM is about 0.2 seconds, which is obviously superior to other methods in recognition speed. The recognition rate of EL-ELM is 87.5% and the template comparison method is 70%.致謝 i
摘要 ii
Abstract iii
目錄 iv
圖目錄 vi
表目錄 viii
Chapter 1 緒論 1
1.1 研究動機 1
1.2 各章內容簡介 2
1.3 論文貢獻 2
Chapter 2 文獻回顧 3
2.1 影像輸入 3
2.1.1 RGB影像 3
2.1.2 灰階影像 4
2.1.3 二值化影像 4
2.2 擷取數字區域影像 5
2.2.1 Sobel 6
2.2.2 形狀因子(Shape Factor) 8
2.3 字元切割 12
2.3.1 連通元件法 13
2.3.2 投影法 15
2.4 數字特徵 16
2.4.1 邊緣統計特徵 17
2.4.2 Histogram of Oriented Gradient特徵 17
2.4.3 討論Sift和Surf特徵 24
2.5 辨識數字 25
2.5.1 邊緣統計 26
2.5.2 樣板比對法(型態學) 29
2.5.3 支持向量機 (Support Vector Machine) 30
2.5.4 放射狀基底函數(Radial Basis Function) 34
Chapter 3 系統架構 38
Chapter 4 Extreme Learning Machines 39
4.1 求解Ax=b 39
4.1.1 Least Square 39
4.1.2 Pseudoinverse 40
4.1.3 Least Square和Pseudoinverse結論 42
4.2 Recursive Least Square 43
4.3 ELM相關學理與數學推導 45
4.3.1 Basic Extreme Learning Machines (Basic ELM) 45
4.3.2 Online Sequential ELM (OS-ELM) 47
4.3.3 Basic Incremental ELM (I-ELM) 48
4.3.4 Enhanced Incremental ELM (EI-ELM) 49
4.4 方法比較 51
Chapter 5 基於ELM水錶數字辨識 53
5.1 水錶辨識基本流程 53
5.2 Basic ELM水錶數字辨識 56
5.2.1 神經元(高斯函數) 56
5.2.2 Basic-ELM訓練樣板和測試樣板的選擇 57
5.3 方法比較 58
5.3.1 樣板比對 60
5.3.2 邊緣統計 62
5.3.3 SVM 65
5.3.4 RBF 67
5.3.5 Basic-ELM 70
5.4 辨識方法和特徵討論 73
Chapter 6 基於ELM之車牌辨識 75
6.1 ELM 81
6.1.1 Basic ELM 83
6.1.2 I-ELM 91
6.1.3 EI-ELM 94
6.2 方法比較 96
6.2.1 樣板比對 96
6.2.2 邊緣統計過程與結果 98
6.2.3 SVM 100
6.2.4 RBF 102
6.3 車牌辨識方法和特徵討論 104
Chapter 7 結論與未來工作 107
7.1 結論 107
7.2 未來工作和方向 107
參考文獻 10