| With the popularity of social media platforms and mobile devices,travellers are used to record what they see and feel during their travels in real time by taking photos and sharing them on the Internet,resulting in a large amount of geographic photo data.Geographic photo data is rich in attribute information,and there are many studies based on the spatial information and textual information related to photos.Therefore,mining the visual information of photographs is of great significance to the study of tourists' behaviour in tourism.In this paper,based on the deep learning algorithm of convolutional neural network,we identify the visual semantic information of images.Facing the problem of complicated recognition results with many visual content elements,we design a quantification method for the visual semantics of tourism photos,and realise the thematic categorisation of photos according to their main visual elements.By mining the visual semantics of individual tourists' photos and classifying the types of tourists,and combining the spatio-temporal information of the photos,the photo-taking behaviour patterns of different types of tourists are analysed and portrayed in terms of both visual semantic features and spatio-temporal behavioural features.In the case study,the city of Xi'an was selected as the study area,and Flickr photos were used as the data source to classify visitor types based on the visual semantic quantification results of visitor photos,and the spatio-temporal information of visitor photos was analyzed comprehensively using methods such as kernel density analysis and 3D visualization of spatio-temporal paths in GIS.The results of the study show that tourists in Xi'an are not a homogeneous group,and that the groups of tourists classified by visual semantics have different characteristics and features in terms of spatial behaviour,as follows.(1)Quantification of the visual semantics of all photographs shows that Xi'an receives the most attention in the category of historical and cultural landscapes,followed by the category of urban space and human life,and at the end by natural landscapes.According to the richness of the photographed landscapes,visitors can be classified into three main types: exclusive,double-themed and multi-themed.(2)The exclusive type of tourists focus on or photograph only the same type of landscape,and their photos show a single-core or double-core spatial density distribution pattern.This type of visitor has a relatively simple travel itinerary and a short average time span,staying for a short time of 1-2 hours at a certain type of landscape.Xi'an has the relatively smallest number of dedicated visitors.(3)The double-thematic type of visitor focuses on two types of landscape or a composite landscape containing two types of visual themes,with a double-core or multi-core pattern of photo spatial density.Their average travel time spans a long period of time,during which they stay in a small range of attractions for a long time and photograph intensively,and their spatio-temporal paths are characterised by simple large-scale patterns and complex small-scale path patterns.(4)The multi-thematic type is the most common type of tourists,whose photos are rich in visual semantics,with random photo-taking actions,and the spatial distribution characteristics of photo kernel density are positively correlated with the hotness characteristics of Xi'an attractions.The spatio-temporal path patterns of their trips are relatively complex and have a wide radiation range,and the duration of their stay in the trip for photography is uncertain.(5)In essence,dedicated tourists and dual-themed tourists are more influenced by their own points of interest,which determine the duration and location of their concentrated photo-taking behaviour during their trips. |
PDF Full Text Request