English  |  正體中文  |  简体中文  |  Items with full text/Total items : 888/888 (100%)
Visitors : 13688551      Online Users : 275
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/363


    Title: 基於三維幾何空間提升對應點關係的運動恢復結構;Correspondence Refinement based on 3D Geometry for Structure from Motion
    Authors: 林彥廷;LIN, YEN-TING
    Contributors: 電機工程研究所
    Keywords: 運動恢復結構;三維重建;特徵點匹配;錯誤傳遞;Structure from Motion;3D reconstruction;features matching;error propagation
    Date: 2017
    Issue Date: 2019-07-17
    Publisher: 電機工程研究所
    Abstract: 藉由二維影像所提供的資訊重建出物體的三維架構一直為值得探討的議題,從古至今所提出的著名方法如運動恢復結構(Structure from motion)、明暗恢復形狀(Shape from shading)…等。傳統的運動恢復結構方法是將攝影機從不同位置拍攝物體所得到的影像中,找出物體特徵點的連續對應關係,從而推算出攝影機姿態,再回推出特徵點於三維空間中之位置。運動恢復結構方法是由影像間的特徵點對應關係重建出三維物體,因此當重建三維物體時,多張影像間二維特徵點對應關係所產生之特徵點軌跡(feature track)是十分重要的。但在傳統的運動恢復結構中,並無特別強調如何串聯多張影像間特徵點的對應關係,常見的方法是以共同影像間的特徵點作為連接的橋樑。這邊我們發現到若當多張影像間的特徵點對應關係中有一小段錯誤的匹配發生卻無被移除時,則此錯誤對應關係會傳遞下去,導致使用此錯誤對應關係估算出錯誤的相機姿態和三維點位置,我們稱此錯誤傳遞關係為連接錯誤傳遞(linking error propagation)。因此我們提出本論文的方法來做改善。本篇論文提出基於三維幾何空間以提升對應點關係並改善當前之運動恢復結構演算法。我們使用三維幾何空間點與點的關係來判斷對應關係間是否可歸類於同條特徵點軌跡,而非單純依影像上特徵點位置來作連接。所以剛開始假設所有影像中的對應關係均為未知狀態,接著使用傳統的運動恢復結構求得兩兩影像間的特徵點對應關係、相機姿態與三維點位置結構,此資訊我們稱為場景圖(scene graph)。接著我們對齊場景圖中提供的三維點結構,若有三維點互相十分接近彼此,我們則可確定這些三維點實際代表物體上同一位置,以此來確定所對應到的特徵點可串聯成同一條特徵點軌跡,同時可減少連接錯誤傳遞的狀況,並提升相機姿態的估測正確性。本論文最後則將所提方法之實驗結果與傳統方法進行比較,以探討其結果差異並提出未來之展望。
    Reconstruct object’s 3D model from the images which are captured by the camera has been discussing for a long time, there have some well-known methods like Structure from Motion, Shape from Shading, …etc.. In the thesis, we focus on the Structure from Motion method. In this method, we have to capture the object’s images from the different viewpoint, then find the feature points’ correspondence over these multi images, based on the correspondence we estimate the camera poses, when we know the camera poses and feature points’ correspondence, finally we can reconstruct the object’s structure.From the Structure from Motion reconstruct steps, we realize that the correspondence of feature points over multi images, which is called “feature tracks”, is very important. In the traditional Structure from Motion, the general way to link the correspondence is to find the same point location at the common image. However we find out that it could occur some problems, when finding the feature track over multi images, if there has a short part of correspondence is wrong but the system treat it as inlier, which means it can not remove properly, so the wrong correspondence will propagate its error, causing the linking error propagation.We propose another way to avoid the linking errors, increase the accuracy of feature tracks, hence estimate robust camera pose. Instead of using the 2D feature points locations to generate the feature tracks over multi images, we consider the 3D-points’ geometry relationship to re-generate the 2D feature points correspondence over multi images. Based on 3D geometry, we are looking forward to avoiding linking errors propagation, so that we can estimate more robust camera poses and 3D-points locations. After that, we will compare the result of our methods with traditional Structure from Motion.In the end, we will summary our methods and make a conclusion.
    Appears in Collections:[電機工程研究所] 學位論文

    Files in This Item:

    File Description SizeFormat
    index.html0KbHTML409View/Open


    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

    隱私權及資訊安全政策

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