Characterizing tourism destination image using photos’ visual content

Abstract

“A picture is worth a thousand words”. Analysis of the visual content of tourist photos is an effective way to explore the image of tourist destinations. With the development of computer deep learning and big data mining technology, identifying the content of massive numbers of tourist photos by convolutional neural network (CNN) approaches breaks through the limitations of manual approaches of identifying photos’ visual information, e.g., small sample size, complex identification process, and results deviation. In this study, 531,629 travel photos of Jiangxi were identified as 365 scenes through deep learning technology. Through the latent Dirichlet allocation (LDA) model, five major tourism topics are found and visualized by map. Then, we explored the spatial and temporal distribution characteristics of different tourism scenes based on hot spot analysis technology and the seasonal evaluation index. Our research shows that the visual content mining on travel photos makes it possible to understand the tourism destination image and to reveal the temporal and spatial heterogeneity of the image, thereby providing an important reference for tourism marketing.