歪斜書背影像切割之研究;A Study For Book Spine Segmentation In Various Orientations

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Computer Science and Information Engineering > Electronic Thesis & Dissertation > Item 987654321/68756

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/68756

Title:	歪斜書背影像切割之研究;A Study For Book Spine Segmentation In Various Orientations
Authors:	張瑞鈞;Chang,Ruei-Jyun
Contributors:	資訊工程學系
Keywords:	書背;切割;book spine;segmentation
Date:	2015-07-22
Issue Date:	2015-09-23 14:24:38 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著科技發展，隨身攜帶的設備像是智慧型手機與平板電腦非常普遍，這些設備通常具備著照相的功能，可以輕易地取得數位影像。當我們身處在一間圖書館內，若要針對某些書籍進行搜尋並且瀏覽其摘要時，目前常見的方式為找到這些書籍所在的類別，並且一本一本地拿下來瀏覽找出想要的書籍。此方法相當麻煩而且浪費了許多寶貴的時間。本研究提出一個以視覺為基礎的系統，利用拍照取得一張包含多個書背的影像，再將影像中的書背一本本地分割並且辨識出來，這項技術可搭配一個書籍資料庫將每本書的各種資訊包含摘要、目錄及作者等告知使用者，藉由影像來協助使用者在偌大的書海中找尋到想要的書籍。本系統利用形態重建法以及L0平滑化技術將影像中的背景以及書背上的文字圖案去除，利用霍夫直線偵測技術將書背邊界擷取出來，並且使用SVM技術補強霍夫直線偵測的缺點，使書背的擷取更為完整。辨識的部分，本論文使用 Speeded Up Robust Features(SURF) 特徵，藉由影像特徵匹配技術確認出正確的書背資訊。我們做了許多實驗比較不同參數的設定以及不同方法的結果，並展示本系統的可靠與有效性。 ;With the development of computer technology, portable devices such as smart phones and tablets have become prevalent. These devices are often equipped with cameras that can get digital images. The common way to seek for certain books in a library is finding the location of these books, and then taking them down one by one to read it. This way is quite troublesome and time-consuming. This study proposes a novel system that segments book spines in images and recognizes them to provide desired information for users. The proposed system takes advantage of morphological image processing and L0 Gradient Minimization image smoothing technology to remove background, graphics and titles on book spines. After that, Hough transform is employed to get boundaries between adjacent books. Furthermore, this work adopts SVM to recover book boundaries missed by Hough transform. As to the part of recognition, we use Speeded Up Robust Features (SURF) to find the correct book spine image in the database. Experiments are conducted to demonstrate the feasibility and effectiveness of the proposed system.
Appears in Collections:	[Graduate Institute of Computer Science and Information Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	281	View/Open

社群 sharing

Loading...