A Study of Book Spine Segmentation and Recognition

• Main Authors: Bo-Yu Li, 李博豫
• Other Authors: Thomas C.Chuang
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/06196874422007427641

碩士 === 萬能科技大學 === 電資研究所 === 104 === In recent years, benefited by improved pattern recognition techniques, applications such as human face recognition, skeleton recognition, gesture recognition, fingerprint recognition, text recognition license plate recognition, character recognition, have been widely applied 0to people's daily lives. A system with great pattern recognition technique brings lot of convenience to people and significantly save time on tasks which mainly relied on manpower in the past ,this thesis discusses the book spine recognition application which applies computer vision and pattern recognition technique . The goal in this study is to reduce manpower use in traditional book inventory system, by using camera on a mobile device. Firstly, a color image is converted to a grayscale image to reduce computation. Then, edges are sharpened with a high-pass filter to highlight book boundaries . Morphological operators are employed to eliminate noises; Hough Transform is then applied to, detect book boundaries. Two types of features, grayscale histogram and SURF features are used for recognition. The proposed system is developed in MATLAB, and several experiments were conducted to verify the efficiency and accuracy of the proposed system.