Identifying books in library using line segment detector and contour clustering

Abstract

In this paper, a new method is presented to automatically identify books based on recognition of call number label on the book spine captured by a smartphone in library. In previous methods for book spine segmentation, Canny edge detector results in curved or broken lines instead of detecting straight lines formed by book spine boundary. Moreover, line detection by Hough transform is prone to interference of texts on the book spine. We address these issues by employing color segmentation with line segment detector to separate every individual book. For call number extraction, characters on the label cannot be easily distinguished from texts like the title and author's name of the book. To solve this problem, we adopt contour clustering to group connected components of characters on the edge map according to their geometrical features. The experimental results demonstrate that our method is effective for the exaction and recognition of books used by existing automated library robotic systems or mobile devices.