Mobile visual search engine and applications

• Main Author: Zhang, Lingzi
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: 2016
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/67485
• Institution: Nanyang Technological University

Mobile visual search is a fast developing technology, which uses images to do search on the mobile phone instead of text. This project aims at developing a mobile visual search system consisting of two key components, which are front-end mobile application and back-end web application. Mobile application is used for customers to recognize objects, for example consumer products, and retrieve dynamic information about the product through network. Web application is designed for clients, who subscribe the services provided by the mobile application, to manage their product database. For the front-end, customers can use the mobile application to capture an image of a product. The image will be transmitted to the server for further processing. Bag of Words model is applied for image matching process. For each query image, SIFT descriptors will be extracted and compared with local image descriptors. Images with highest scores will be taken as candidates and returned. Geometric verification is to be done to select the best candidate who gives the highest consistency from different viewpoints. For the back-end, clients can use the website as a platform to upload and modify images and details of their products stored in the database. The website functionality is enhanced by allowing different clients to define unique attributes for their products when creating the tables in the database. A personalized database will be built for each client. Meanwhile, the website interface is designed and polished in an appealing style. To optimize the user experience, data mining technique is utilized to provide services to customers and clients. This project focuses on applying association rule mining on transaction data to explore relevance among different products. For online reporting, customers can look through the recommendation page on the mobile app to find out products related to the searched product. For offline reporting, clients can check the reporting page on the website, which gives them real-time feedback on sales or stock and suggestions on sales plan based on association rules.