Mobile visual recognition landmark (design of mobile visual search system)

• Main Author: Tang, Niting
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/64153
• Institution: Nanyang Technological University

Mobile Visual Search system is an integrated system consisting of web client side and mobile client side. Mobile visual search performs search on the mobile through image recognition. By using Bag-of-words model, images are represented and classified by local features descriptor. Scalable vocabulary tree with inverted index is applied to the image database to improve retrieval efficiency and quality of feature matching. Features of query image and database images are checked for matches, and database images with most matches are selected for geometric verification. Geometric verification is a technique to verify the spatial consistency between query image and selected database images. After that, matched images with associated information are extracted and return to the mobile. Web application was developed to allow subscribers to manage and upload images when they would like to compose the image database for mobile visual search. Images and related information could be uploaded as a batch or increment. To complement the web application to an integrated web site, registration and secure access system were carried out. Server was built to store uploaded images and interpret web content scripts, and MySQL database was created to store account information and related information. Integrated mobile visual system was generated by sharing the server and database of web application and mobile application. Images uploaded via web application would form image feature database for mobile visual search and image extraction. Also, related information inserted into MySQL database via web application could be extracted by mobile application. Therefore, an integrated mobile visual search system was generated. Testing on web application and integrated system was conducted by using book cover and information database. Results verified the capability of all web application functionalities and the overall system.