Content and context information fusion for mobile landmark recognition

This paper proposes a new information fusion approach that employs two information components for mobile landmark recognition, which includes: content analysis and context analysis. Existing landmark recognition works are mainly based on PC platform, which uses content analysis alone for recognition, and thus has a large computation cost and cannot satisfy mobile users´ fast response time requirements. Therefore in this paper, a mobile platform-based landmark recognition system is developed. The bag-of-words method is employed and further improved as content analysis according to the specific features of landmark images. Context analysis based on location (obtained by global positioning system) and direction (obtained by digital compass) information is then integrated with improved BoW as a new information fusion approach for mobile landmark recognition. Finally, the experiment results conducted on the NTU landmark database show that the proposed approach is effective for mobile landmark recognition.