A lot of street view services, which present views of urban landscapes, have recently appeared. The conventional method for making street views requires on-vehicle cameras. We propose a new method, which relies on people who voluntarily take photos of an urban landscape. We have developed a system called QyoroView. The system receives photos from users, adjusts the photos’ position and orientation, and finally synthesizes them to generate a street view. We conducted two experiments in which the subjects generated a street view using our system. We also observed and interviewed the subjects who participated in order to learn their impression of the system.
There are several approaches to producing street views. The basic style of this content is a hand-crafted panoramic image (Chen, 1995). An enormous amount of effort is required to construct wide-area panoramic images (Hartono et al., 2006). Using this approach, it is possible to produce street views of some spots, but it is nearly impossible to produce views along streets.
Past studies proposed several approaches to produce street views automatically using on-vehicle equipment. A common technique for collecting panoramic images of many spots along streets has been to use on-vehicle omnidirectional cameras (Koizumi, & Ishiguro, 2005; maps.google.com/help/maps/streetview; preview.local.live.com). A previous study used on-vehicle laser scanners to obtain 3D models of buildings along streets (Fruh, & Zakhor, 2001). Another previous study used on-vehicle cameras mounted to the side of a car to capture panoramic images of buildings along streets (Roman, Garg, & Levoy, 2004). These approaches require special equipment and long-distance driving. Thus, it is costly to produce wide-area street views.
‘Photo Tourism’ is a system that can make street views from a collection of photos that are freely available in photo sharing sites (Snavely, Seitz, & Szeliski, 2006; labs.live.com/photosynth). So, it requires no equipment or labor to produce street views. However, a sufficient number of photos recording the same location are needed to concatenate photos using image processing technology. Many sightseers take photos at tourist attractions and some of them upload their photos to photo sharing sites, so this approach is applicable to the production of street views of such places. It is, however, not a good idea to apply the approach to the production of street views of featureless urban areas due to the lack of motivation for users to take photos.
‘OpenStreetMap’ is a UGC system that makes freely editable maps. The map data in the system has mostly been built by users who upload data from their phones or PDAs equipped with GPS devices (openstreetmap.com)]. Recently, Yahoo! allowed OpenStreetMap to use their vertical aerial imagery. So, users can create vector-based maps using online editing tools that can overlay aerial images. The system produces only vector-based maps of urban streets.
Key Terms in this Chapter
Vector Map: Digital data of maps that is consisted of X-Y coordinates. If you draw lines and polygons with these coordinates, you can create maps
User-Generated Content: Media content created by end users
Border Vector: Digital data of Border line between a road and a City block
Continuous Capture Method: A method of uploading several photos at once
One-Shot Capture Method: A method of uploading only a single photo
Photo Map: A map onto which are placed photos uploaded from end users
Halfway Vector: Digital data of Center line of a road
Street View: Panoramic images of urban landscapes that are made from movies or a collection of photos taken along the streets of a city