Visual search in mobile business

• Main Author: Phyo, Ye Minn
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems
• Online Access: http://hdl.handle.net/10356/65757
• Institution: Nanyang Technological University

With the advancement of technology, a lot of people have been using smart phones in their daily lives and eventually those became part of their lifestyle. Indeed, many people can easily capture the images with their smart phones’ cameras and post them into social networks like Facebook and Instagram over a WI-FI or 4G network. Besides, many people have an interest on the information of the food they are having everyday when they are dining out with their loved ones on weekends or whenever at their convenience. Nowadays, there are a lot of research and development software for image recognition. One good example is ‘Google Goggles’. But there is only a few to none image recognition software available for food menus in restaurants. This project aims to assist smart phone users to gather information about the food menus that they can easily capture with their smart phones’ cameras. Various image recognition techniques will be discussed in this project. Moreover, comparison between food menus images under different circumstances will also be discussed. A distinct limitation in capturing food menus is the unwanted features in the surrounding of menus and occlusions made by the users. And those will result in noise in image matching process and affect the matching accuracy. A powerful technique called Geometric Verification (GV) will be introduced in feature matching to eliminate those issues. It is a relatively effective method and can increase the recognition accuracy at a range of 5 to 10%. One drawback of GV is that it will take longer processing time in feature matching process. A set of experiments was done to compare and contrast the image processing with and without GV. After that, the ways of improving matching accuracy in capturing food menus and optimization of dataset for future use will also be discussed.