Utilizing internet photos for indoor mapping and localization - opportunities and challenges

Along with the booming popularity of photo sharing services, an ever-increasing number of photos are shared and made publicly available on the Internet. This brings new opportunities for image-based applications and services. In this paper, we explore the potential of using Internet photo collections for building 3D models of indoor environments and providing indoor mapping and localization using these models. We choose several representative places of interest in order to analyze their Internet photo collections. Photo sets have been collected from two wellknown public attractions to generate 3D models using Structure from Motion (SfM) techniques. We further design several metrics to evaluate the usability of these models for indoor mapping and localization. The results show that Internet photo collections can serve the purpose of indoor mapping and localization in certain types of indoor spaces.