English  |  正體中文  |  简体中文  |  Items with full text/Total items : 27320/39164
Visitors : 2475026      Online Users : 32
RC Version 4.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Adv. Search
LoginUploadHelpAboutAdminister

Please use this identifier to cite or link to this item: http://ntour.ntou.edu.tw:8080/ir/handle/987654321/35767

Title: 在遮蔽情況下使用稀疏編碼之動作分析技術
Action Recognition under Occusions Using Sparse Coding
Authors: Kai-Ting Chuang
莊凱婷
Contributors: NTOU:Department of Computer Science and Engineering
國立臺灣海洋大學:資訊工程學系
Keywords: 稀疏編碼;稀疏表示;遮蔽;行為偵測
Sparse Coding;Sparse Representation;Occlusion;detect the daily event
Date: 2012
Issue Date: 2013-10-07T02:58:53Z
Abstract: 近年來行為分析在電腦視覺中,是很熱門的一項領域且廣泛應用於智慧型監控方面,例如:異常行為偵測、環境安全監控、遠距醫療監控…等。然而以往的行為分析皆為單人行為居多,因此,本篇論文將以日常生活中常見的雙人行為 (打招呼、握手、走路…等)做分析。由於雙人行為會有遮蔽的情況發生,如何提出一個表示動作行為的方法,正是我們所面臨的挑戰。 本文提出了一種基於稀疏信號表示方法,以解決視頻中的人體動作識別的問題。對於每一個動作,一組冗餘的基礎上(字典)是通過求解稀疏優化問題。透過學習每一個動作字典,每個字典能夠有效地代表一個特定的動作。 系統主要可由三部分所組成,分別是偵測特徵區域、特徵抽取與利用稀疏編碼(Sparse Coding)來進行行為分析。首先,以前景偵測的技術抽取待分析的前景人物並針對前景偵測出人,接著對於偵測區域抽取所需的R Transform特徵與方向梯度直方圖,並以特徵來表示一序列動作,利用稀疏編碼將各個動作串聯起,來描述動作之間的轉移特性,並結合稀疏表示分類與漢明距離分類來分析並獲得行為動作。 由實驗結果得知,我們所提出的方法,對於在遮蔽情況下,雙人在日常生活中的行為分析與分類,有相當不錯的效果。
Recently, behavior analysis has become an important task in computer vision, and is expected to enable many applications, such as intelligent video retrieval, human interaction system, and so on. Most existing visual behavior researches put focus on analyzing single behavior. Therefore, we present a general model for several highlighted daily behaviors (e.g. greeting, shaking hands, and walking), which works efficiently and effectively although occlusions occur. In this thesis, we based on sparse representation to solve the problem of human behavior recognition in video. For each behavior, a overcomplete dictionary is solving the sparse optimization problems. Through the learning of every behavior dictionary, each dictionary can effectively represent a specific behavior. The proposed system consists of three components: foreground region extraction, feature extraction, and Sparse Coding to behavior analysis. First of all, foreground objects and people detection information are extracted by using Gaussian mixture model and Connected Component. Second, features are extracted by using R Transform and Histogram of Oriented Gradients. Finally, we employ a Sparse Representation-based Classification and Hamming Distance Classification to infer the problem of behavior analysis.
URI: http://ethesys.lib.ntou.edu.tw/cdrfb3/record/#G0019957016
http://ntour.ntou.edu.tw/handle/987654321/35767
Appears in Collections:[資訊工程學系] 博碩士論文

Files in This Item:

File Description SizeFormat
index.html0KbHTML133View/Open


All items in NTOUR are protected by copyright, with all rights reserved.

 


著作權政策宣告: 本網站之內容為國立臺灣海洋大學所收錄之機構典藏,無償提供學術研究與公眾教育等公益性使用,請合理使用本網站之內容,以尊重著作權人之權益。
網站維護: 海大圖資處 圖書系統組
DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback