臺灣博碩士論文加值系統

當人們旅遊時，通常會在旅途當中用照片記錄下沿途優美的景色或造型時尚雄偉的建築物，並將這些照片上傳到社群媒體上分享。現今，在每一秒都會有大量的照片上傳到社交網站上，像是Flickr，Instagram，Panoramiｏ等。歸功於智慧型行動裝置的普及，這些上傳到社交網站上的照片通常含後設資料(meta data)，像是標籤，時間，地理位置等。基於這些相片和其後設資料，我們提出一個旅遊景點推薦系統。藉由此系統，使用者可以輕易的找出他們所偏好的旅遊景點。我們的系統首先需要使用者上傳一張照片(query photo)，利用這張照片的色彩(color)及紋理(texture)從影像資料庫中找出相似的影像，接著使用者可以透過我們所提供的使用者互動介面，從這些相似的影像中選擇「喜歡」或「不喜歡」的照片作為回饋(feedback)。透過這些回饋相片的後設資料，系統可以更準確的預測使用者偏好的旅遊景點，並將使用者「喜歡」相片附近拍攝的景點一併呈現在介面上，供使用者參考。

When people go traveling, they usually take photos throughout their journeys, recording magnificent scenery or buildings they have visited and sharing with their friends by social media. Nowadays there is vast quantity of this kind of photos being uploaded to social networking websites, such as Flickr, Instagram, Panoramio, etc., in every single second. Thanks to the popularization of mobile devices with GPS chip, these photos always contain rich meta data, including not only tags and time but also geo-locations. With this rich information, we propose a tourist attraction recommendation system. Users can easily find the preferred tourist attractions by our system. At the beginning, a query photo uploaded by a user is required. Our recommendation system retrieves similar photos from the geo-tagged photo database according to the similarity of color and texture first. Then, users can select “like” and “dislike” photos as the feedback among them. Through tags and geographical features of neighboring places of these photos selected by users, we can more precisely predict what kind of photos preferred by users and filter out those photos not preferred by users. Moreover, those photos taken near the geo-location of each “like” photo are also shown on the interface for users’ reference.

Chapter 1 Introduction......................1
1.1 Motivation..............................1
1.2 Related Work............................3
1.3 The Proposed System.....................7
Chapter 2 The Proposed Tourist Attraction Recommendation System......................................9
2.1 Preprocessing...........................10
2.1.1 Visual Features Extraction............10
2.1.2 Tag Scoring...........................15
2.1.3 Geographical Features Extraction......17
2.2 Visual Feature Matching.................20
2.3 User Feedback...........................22
2.3.1 Refinement Using Tags.................22
2.3.2 Refinement Using Geographical Features..25
Chapter 3 User Interface Introduction.......28
Chapter 4 Experimental Results..............31
Chapter 5 Conclusion and Future Work........36
References..................................38

[1] J. Hays and A. A. Efros, “IM2GPS: estimating geographic information from a single image,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pp. 1-8, June 2008.
[2] W. He, R. Li, Z. Wu, J. Hu, Y. Liu, “Generating landmark overviews with geo-tagged web photos,” IEEE International Conference on Systems, Man, and Cybernetics, pp. 2644-2649, October 2012.
[3] L. W. Renninger and J. Malik, “When is scene recognition just texture recognition?” Vision Research, Volume 44, Issue 19, pp. 2301–2311, September 2004.
[4] T. Leung and J. Malik, “Representing and recognizing the visual appearance of materials using 3-dimensional textons,” International Journal of Computer Vision, Volume 43, Issue 1, pp. 29-44, June 2001.
[5] D. Comaniciu and P. Meer, “Mean shift: a robust approach toward feature space analysis,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Volume 24, Issue 5, pp. 603–619, May 2002.
[6] Flickr.com. http://www.flickr.com/.
[7] http://www.flickr.com/services/api/flickr.photos.search.html.
[8] GeoNames. http://www.geonames.org/.
[9] M. Stricker and M. Orengo, “Similarity of color images,” Proceedings of the SPIE Conference on Storage and Retrieval for Image and Video Databases III, Volume 2420, pp. 381­392, February 1995.
[10] C. D. Manning, P. Raghavan and H. Schütze, “Introduction to Information Retrieval”, pp. 158-160, Cambridge University Press. 2008.