貝蘇章臺灣大學：電信工程學研究所陳勁元Chen, Chin-YuanChin-YuanChen2007-11-272018-07-052007-11-272018-07-052007http://ntur.lib.ntu.edu.tw//handle/246246/58731視訊在今日已是隨手可得。自動化理解視訊的關鍵之一在於分析其中的動作。在過去的二十年間，有許多相關這個主題的研究。 首先，關於回顧前人研究中關於動作分析的一般步驟，包含動作偵測、動作分段、物體分類。接著，提到一些與動作分析相關的應用，例如視訊監控及個人辨識。整理並呈現與這些應用相關的處理技巧，諸如物體追蹤與行為了解。 然後，引入一個可以直接了解視訊中蘊含行為的方法。我們集中在使用光流以達成在視訊中辨識動作的目標。這個方法運用時間及畫面所成的體積找出待測視訊間的關係。所有的處理從最小的基本單位開始，這個基本單位稱為時空小塊。藉由釐清時空小塊之間的種種性質，可以找出由時空小塊構成的大型待測視訊間的關係。實驗結果證明這個方法是有效的。其基本概念既直接且簡單，而且不受待測視訊中動作物體的外觀影響，但是在進行實驗時必須詳加考慮此法的高運算複雜度。Videos are all around us. One of the key to understand a video automatically is to analyze to motions within it. In the past two decades, researches have been conducted against this topic. First, the general procedure of motion analysis is reviewed in the related work, including motion detection, motion segmentation, and object classification. Then, some applications of motion analysis are mentioned, such as video surveillance and personal identification. Techniques related to these applications are summarized, such as object tracking and behavior understanding. Then, a direct way to understand the behavior of a video is introduced. We focus on this optical-flow based approach to recognize action in video sequences. This method uses space-time volumes to perform correlation between templates. All the process of this approach start with the basic unit, space-time patches. With the well developed properties of these space-time patches, the correlation between two large templates can be practiced. Some results from experiments have proved the validity of this method. The concept of this method is direct and simple, and it is irrelevant to the appearance of the moving object, but the implementation of experiments should be designed with the consideration of high computation complexity.Chapter 1 Introduction 1 1.1 In the age of media explosion 1 1.2 Aimed issue 2 Chapter 2 Related works 3 2.1 Motion detection 3 2.1.1 Background update 3 2.2 Motion segmentation 4 2.2.1 Background subtraction 4 2.2.2 Temporal difference 5 2.2.3 Optical flow 5 2.2.4 Hybrid method 5 2.3 Object classification 5 2.3.1 Shape-based classification 6 2.3.2 Motion-based classification 6 2.3.3 Other way 7 2.4 Recent popular application 7 2.4.1 Object Tracking 8 2.4.1.1 Region-based tracking 9 2.4.1.2 Active contour-based tracking 10 2.4.1.3 Feature-based tracking 11 2.4.1.4 Model-based tracking 13 2.4.2 Understanding and description of behaviors 21 2.4.3 Personal identification for visual surveillance 26 CHAPTER 3 Motion Correlation Based on Space-Time Optical-flow 34 3.1 Properties of a space-time intensity patch 35 3.2 Consistency between two ST-patches 37 3.3 Handling spatial-temporal ambiguities 39 3.4 Continuous rank-increase measure Δr 45 3.5 Correlating Space-Time Video Templates 47 3.6 Considerations for experiment 48 3.7 Experiment results 52 3.8 Conclusion 58 CHAPTER 4 Alternative Method for Motion Correlation Based on Space-Time Optical-flow 60 4.1 Properties of a space-time intensity patch 60 4.2 Consistency between two ST-patches 61 4.3 Correlating Space-Time Video Templates 62 4.4 Comparisons with the rank-increase method 63 4.5 Experimental results 64 4.6 Conclusion 70 CHAPTER 5 Future work 72 Reference 74en-US利用時空光流比對來做動作辨識Action Recognitionthesis