Research And Implementation About The Museum Navigation System Based On Image Retrieval Teachnology

In traditional museums, sightseers are usually guided by commentators. Theyspend extra money on navigation service, but cannot freely enjoy visiting. Currently,most of the museum navigation systems are based on RFID technology. Since visitorscan only receive voice information about exhibits passively, the User Experience (UE)of such systems is not satisfactory. In recent years, the emerging touching screentechnology and mobile APPs greatly improve the UE of museum navigation systems.However, visitors cannot still search information about exhibits during visiting. In thispaper, the museum navigation system based on image retrieval is proposed. Byfollowing the museum’s official number on the public platform of Wechat APP or login the museum’s official website, users can take photos of exhibits for searchingrelated information. Our proposed navigation system can offer much more excellentUE.The main contributions of this paper are as follows:(1) Database’s creation: By taking photos in the Guangdong Provincial Museum,Guangzhou Museum, Guangdong Museum of Art and other, our own test database iscreated.(2) we propose and implement a novel image retrieval algorithm based onmulti-feature(color histogram, SIFT, SURF). The algorithm combines the advantagesof color histogram, SIFT and SURF features.The experimental results show that theperformance of the proposed algorithm is better than that of the algorithm use onlycolor histogram feature, SIFT feature or SURF feature. (3) Propose and implemente a positioning algorithm of the main area based onfeature points distribution of the exhibits images to remove the background of exhibitsimages properly.(4) Design and implement the prototype of the museum navigation system whichis based on image retrieval, and deploy it on the public platform of Wechat APP andthe website.Because our system employ exhibit images as the basis of query, it’s easy forusing. By the aid of the platform of Wechat APP and the website, the trouble ofsoftware’s downloading is avoided. Due to a layered K-means method which is usedto construct the inverted file index and visual-characteristics-based vocabulary tree, itis possible to perform query on the large-scale image characteristics database.And oursystem has good scalability, the administrator can maintain multimedia informationdatabase of products and the image library at any time.