指導教授：洪一平臺灣大學：電機工程學研究所廖偉傑Liao, Wei-JieWei-JieLiao2014-11-282018-07-062014-11-282018-07-062014http://ntur.lib.ntu.edu.tw//handle/246246/262907在眾多視訊監控的應用中，前景偵測和影子的消除一直是一個很重要的議題，我們對於前景偵測和影子消除提出了一個新的架構，我們的架構包含兩個五要的部份分別是前景偵測和影子偵測。我們提出時空背景截取器來進行前景偵測，時空背景截取器主要包含背景截取和背景邊緣截取兩部份，時空背景截取器在動態背景和瞬間光線變化有不錯的表現。在影子消除的部份我們選擇色度、物理性質和紋理三種特徵來當作判斷的資訊，我們利用隨機森林分類器和所選擇的特徵學習出適合各個場景的影子偵測模型。我們使用建出來的影子偵測模型針對時空背景截取器的前景結果進行影子消除。除此之外我們和目前較常使用的前景偵測和影子消除的方法進行比較。Cast shadows detection and removal is indispensable in the object detection to many surveillance applications. In this paper, we present a novel framework for removing cast shadow of moving objects. Two main components, moving objects detector and redundant shadow remover, are integrated. For moving objects, we adopt the spatiotemporal background extractor (SBE) to detect the moving objects which is comprised of the background extractor (BE) and the background gradient extractor (BGE). SBE features the object detection in the dynamic background and the sudden lighting changes environment. For shadow removal, we use the classifier, Random Forest, to learn the shadow features, which are chromaticity, physical properties, and texture. Then, we remove the shadow from the result of SBE with the shadow classifier. The proposed method can effectively detect the moving objects and remove the shadow effect. Furthermore, we demonstrate the performance of our method compared with some state-of-the-art techniques of object detection and shadow removal.口試委員會審定書 # 誌謝 i 中文摘要 ii ABSTRACT iii CONTENTS iv LIST OF FIGURES vi LIST OF TABLES viii Chapter 1 Introduction 1 1.1 Motivation 1 1.2 Background Knowledge 1 Chapter 2 Related Work 3 2.1 Background Modeling 3 2.2 Shadow Removal 5 2.3 Feature of Shadow 8 2.4 Classifier 10 Chapter 3 Methodology 13 3.1 System Architecture 13 3.2 Spatiotemporal Background Extraction 14 3.2.1 Background Extractor (BE) 15 3.2.2 Background Gradient Extractor (BGE) 19 3.3 Shadow Removal Classifier 21 3.3.1 Feature Extraction 22 3.3.2 Classifier Training 24 Chapter 4 Experiments 26 4.1 Experimental Result Analysis of Spatiotemporal Background Extractor 27 4.2 Experimental Result Analysis of Shadow Removal with Single Image 29 4.2.1 Combined Strategy between Different Features 31 4.2.2 Experimental Result of Different Classifier 32 4.3 Experimental Result of the Proposed Method 34 Chapter 5 Conclusions and Future Works 37 REFERENCE 391899273 bytesapplication/pdf論文公開時間：2019/08/25論文使用權限：同意無償授權前景物偵測影子消除支持向量機隨機森林thesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/262907/1/ntu-103-R01922106-1.pdf