English  |  正體中文  |  简体中文  |  Items with full text/Total items : 888/888 (100%)
Visitors : 13688747      Online Users : 265
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version
    Please use this identifier to cite or link to this item: http://ccur.lib.ccu.edu.tw/handle/A095B0000Q/356

    Title: 基於多重協同攝影機視訊之人體 3D 姿態監控與異常行為辨識;3D Human Posture Monitoring and Abnormal Behavior Identification Based on Coordinated Cameras
    Authors: 謝政翰;HSIEH, CHENG-HAN
    Contributors: 電機工程研究所
    Keywords: 協同攝影機;人體3D姿態監控;異常行為辨識;Coordinated Cameras;3D Human Posture Monitoring;Abnormal Behavior Identification
    Date: 2017
    Issue Date: 2019-07-17
    Publisher: 電機工程研究所
    Abstract: 本論文以常見的 IPCam 擷取視訊影像處理2D及3D 特徵和人工智慧深度學習的應用著手,與近年來常見的深度攝影機應用方式不同,並無取得空間中之深度資訊進行處理辨識,而是從監控場景中人體的異常行為辨識為核心要件,期望在 2D 影像中套入 3D 的特徵訊息,並結合深度學習技術獲得較精準之辨識準確率,如此便可應用於居家安全監控、人機互動或異常行為辨識等大方向應用。 本論文所提出之基於多重協同攝影機視訊之人體 3D 姿態監控與異常行為辨識之演算法包含三個主要步驟,首先是人體 3D 骨架模型建立,利用 3D 骨架特徵取得人體於空間中的信息。第二為人體姿態辨識,以人體輪廓特徵結合深度學習技術進行姿態分類與辨識。最後才以帶有 3D 骨架之動作序列搭配深度學習技術進行異常行為辨識,因此本論文即是利用 2D 影像特徵與 3D 空間特徵的方式進行人體動作的分析與辨識。 在深度學習方面,本論文將採用卷積神經網絡 (CNN) 來對視訊影像進行處理,其對於特徵輸入為影像形式時的應用有其規律的步驟流程,利於影像特徵的轉換並保留其原始資訊,透過實驗結果,可以得知 CNN 對於影像處理的功效良好,較其餘深度學習之技術來的更適合應用在影像特徵當中,因此常見 CNN應用在人體偵測或動作辨識當中。
    In this study, we use deep learning to process 2D and 3D image features, which is extracted from IPCam. Application is different from the depth camera, It’s doesn’t need to extract the depth information in the image, we identify the abnormal behavior by adding the 3D features of the human gesture in the monitoring scene, and add 3D features to the image, and combined with the deep leaning makes the accuracy more accurate. This can be used in home care, human-computer interface or abnormal behavior identification.The algorithm consists of three main steps proposed in this study. First, the 3D skeleton model is built on the human body, and 3D skeleton features are used to obtain information about the human body in space. The second is to identify the human posture, classification and identification of human posture using human contour feature combined with deep learning. Finally, the action sequences with 3D skeleton and depth learning are used to identify the abnormal behavior.In deep learning, we use convolution neural network (CNN) to deal with video images, and it have regular process when the input feature is the image, which will facilitate the image feature conversion and retain their original information. Through the experimental results, we can see that CNN for image processing is better than the other deep learning technology.
    Appears in Collections:[電機工程研究所] 學位論文

    Files in This Item:

    File Description SizeFormat

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

    版權聲明 © 國立中正大學圖書館網頁內容著作權屬國立中正大學圖書館


    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library IR team Copyright ©   - Feedback