以字庫法為基礎之圖書影像盤點系統

An Image-based Bookstore Inventory Check System Using Bag-of-Words

周秉宜(Bin-Yi Chou)

指導教授 ： 劉震昌

國立暨南國際大學/科技學院/資訊工程學系/碩士(2010年)

https://doi.org/10.6837/NCNU.2010.00214

摘要

在圖書庫存管理中，盤點作業相當耗費時間及人力，無論是圖書館或一般書店都必須完成此項沉重的業務來掌握書目及藏書檔的正確性與完整性，然而此項作業需在事前作好妥善的規劃以確保盤點品質及速度，所以作業週期較為長，也因此管理者常常無法及時地控管書目資料的品質。
本論文提出以影像作為書籍資訊的圖書管理方法，上架完成時利用架設於書櫃附近的 PTZ 網路攝影機來取得書架上的書籍影像，而這些書籍影像經過一個以有限狀態機為基礎的方法進而將書籍的書背影像切割出來，然後對其作 SIFT 特徵的擷取，並與書籍相關資訊 (如：書籍寬度、書籍版本、國際標準書號、書籍位置、書籍狀態…等) 儲存於資料庫中。盤點時利用相同方法取得書籍影像、切割出書背影像、擷取特徵，與上架完成時儲存的圖書影像資訊利用影像搜尋技術進行比對。然而龐大的圖書影像資料仍需耗時比對，所以使用字庫法 (Bag-of-Words) 來加快比對的速度，盤點完成後，取得架上的書籍資訊。最後，根據盤點結果更新資料庫所有的書籍相關資訊，及時地掌握目前藏書檔的正確性。實驗後使用窮舉法搜尋書背可以達到九成左右的辨識率，使用字庫法改進後每本書的搜尋時間可以降低到 0.05 秒左右。

關鍵字

書店管理 ； 盤點 ； 影像搜尋 ； 字庫法

並列摘要

Book inventory check in libraries or bookstores is a time-consuming and labor-intensive work because of the enormous number of books, but it is necessary for the managers to keep track of the latest information about the books in stock. In addition, careful planning in advance is key to the quality and speed of the inventory check. Thus, the inventory check period is usually very long such that the managers cannot obtain correct information of book inventory timely.
In this thesis, an image-based approach is proposed for automatic book inventory management. Bookshelf images are captured by using a number of PTZ web cameras distributed around the bookshelves. Then, a model-based approach is applied to segment bookshelf images into individual book-spine images in which the SIFT features are extracted. The SIFT features along with other information such as the width, the version, ISBN, and the position of the book are stored in the database. To run the inventory check, the same procedure is applied to extract SIFT features of the query book-spine image, then the most similar book-spine image in the database is found via exhaustive feature matching. However, this exhaustive search method is time-consuming because of the enormous number of database images. The concept of Bag-of-Words (BOF) is used to improve the efficiency of search. Experimental results show that the recognition performance using exhaustive search is about 90% and search time using BOF is about 0.05s for each book.

並列關鍵字

bookstore management ； inventory check ； image search ； Bag-of-Words

參考文獻

[1] A. Gionis, P. Indyk, and R. Motwani, “Similarity Search in High Dimensions via Hashing,” in Proceedings of the Twenty-fifth International Conference on Very Large Databases, pp. 518-529, 1999.

[2] A. Koschan and M. Abidi, “Detection and Classification of Edges in Color Images,” IEEE on Signal Processing Magazine, Vol. 22, pp.64-73, 2005.

[3] D. G. Lowe, “Object Recognition from Local Scale-Invariant Features,” in Proceedings of the Seventh IEEE International Conference on Computer Vision, Vol. 2, pp. 1150-1157, 1999.

[4] D. Nister and H. Stewenius, “Scalable Recognition with a Vocabulary Tree,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, Vol. 2, pp. 2161-2168, 2006.

[5] E. Levina and P. Bickel, “The Earth Mover’s Distance is the Mallows distance: Some Insights from Statistics,” in Proceedings of the Eighth IEEE International Conference on Computer Vision, Vol. 2, pp. 251-256, 2001.