Image Feature Point Matching on Landmark Image Retrieval Applications
| Autor: | Ning, Ouyang, 歐陽寧 |
|---|---|
| Rok vydání: | 2013 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 101 With the development of technology, portable devices such as smart phones and tablets have become prevalent. These devices are often equipped with Global Positioning System (GPS) and are able to locate its current approximate location. Such information is very useful for travelers. However, when people are travelling, the GPS readings alone could not accurately give the exact landmark information to the device carriers. Sometimes the travelers would observe a landmark from a distance. In addition, multiple landmarks could be observed at the same place. Under such circumstances, if the travelers wish to retrieve the landmark information, they would need to look up all the landmarks nearby and try to get the information they are looking for. Since the portable devices are also equipped with cameras, we propose an application that can utilize the images captured by the cameras and the location information acquired by the GPS readings and rapidly return the desired landmark information to the users. The user can easily take a photo and touch a button to search for the correct landmark information from famous travel websites. The proposed system takes advantage of the Google place API and gets the names and types of the nearby places of the current GPS reading. The names are filtered first to get the landmark candidates. Then the keywords of these landmark candidates are used to query the image database of the public search engines. The returned images are then confirmed by the system using local feature matching techniques. In the thesis, we use Speeded Up Robust Feature (SURF) and Oriented FAST and Rotated BRIEF (ORB) features. The system analyzes these images by feature matching to confirm the correct image of the landmark. We also use FLANN library to construct randomized k-d tree. After confirming the landmark name, the system would query the famous travel websites using the correct landmark name and return the search results to the users. We have performed experiments to compare the performance of different features and parameter settings. We have also demonstrated the feasibility and effectiveness of the proposed application system. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|