Typically, the look of a hallway will not change completely. Visual alterations to posters and door signs, for example, only tend to affect a limited area of a particular location. New photographs enable NAVVIS to update its map whilst simultaneously allowing a user to localise. The system gets better and better over time.
Researchers from the Technische Universität München (TUM) have developed a positioning system that enables people to navigate areas beyond the reach of Global Positioning System (GPS) signals. NAVVIS, a tool that uses mostly visual information, allows smartphone users to navigate indoor structures impervious to satellite signals. In addition to navigation, the TUM team is confident that their system can achieve useful applications within the field of augmented reality.
The first step in the NAVVIS process is to catalogue a building using a specially-designed mapping trolley. The trolley, which includes two laser scanners, single lens reflex (SLR) cameras and a panoramic camera, is pushed along the corridors of the building in question. As the trolley passes through the building, its lasers scan dimensions both horizontally and vertically in order to create a virtual map using three-dimensional point clouds. The researchers then use software to transpose the captured photographs onto the pixels of this map. The result is a realistic 3D view of the building’s interior.
Once a building has been mapped, a smartphone app developed by the researchers acts as the navigational tool. The app uses a phone’s camera to take a snapshot of its environment, and matches this photograph with the other images stored in system’s database. From this information, NAVVIS can inform the user of their exact location and
tell them in which direction they are pointed. The system then uses arrows and a 3D view to guide the user to his or her desired destination.
I spoke to project leader Georg Schroth from TUM’s Institute for Media Technology to find out more about this novel visual localisation system. I began by asking Schroth whether or not NAVVIS is as accurate as GPS.
"That very much depends on the type of GPS signal that you are using," he replied. "NAVVIS is accurate to one metre of the user’s location which makes it ideal for use within environments such as universities and museums. It also offers information regarding orientation, which sets it apart from other navigational systems. In terms of orientation, our system offers an accuracy of between 10 and 20 degrees."
GPS signals are unsuitable for indoor navigation and whilst Wi-Fi offers one possible solution to this problem, its accuracy is fairly limited. The visual-based nature of NAVVIS, on the other hand, provides a high degree of accuracy without the need for any significant investment in infrastructure.
NAVVIS uses both vertical and horizontal laser scans during mapping which allows the surrounding environment to be displayed as a three-dimensional point cloud
"GPS doesn’t suit interior scenarios because its signals are overly attenuated," explained Schroth. "There is no way to overcome this problem. For this reason, companies such as Nokia, Samsung, Google and Apple have been developing indoor localisation systems based on wireless technology. Wi-Fi has several advantages. Every smartphone is able to receive wireless signals and many buildings contain Wi-Fi hotspots. This approach, however, has one major drawback. Wi-Fi localisation is impaired by factors such as the number of people walking around an environment. Realistically, therefore, with a typical Wi-Fi localisation system, it is only possible to achieve an accuracy of approximately room level. For a greater degree of accuracy, you would require special circumstances such as having a large number of Wi-Fi hotspots in range.
"The advantage of a visual localisation system is that you don’t have to make any infrastructural investment. Wi-Fi hotspots and Bluetooth beacons are not necessary. You simply need to use the trolley to map the environment in question. The localisation system can be ready for use the very next day."
I went on to ask Schroth whether or not this process is time consuming or complicated. As he explained, the task can be performed within a single day and requires no expert knowledge.
"In order to scan a building, you must simply roll the mapping trolley around its corridors," he said. "The trolley has laser scanners and a panoramic camera, and can be pushed around at walking speed. Of course, the time that this takes varies according to the size of the building in question. Mapping a large building might take a day but smaller buildings can be mapped within a couple of hours. This is not a complicated process; it doesn’t take an expert to map a building."
One potential caveat for visual localisation systems is the fact that buildings rarely stay the same for prolonged periods of time. Before long, the photographs taken during a building’s initial mapping might not correspond with the look of a building at a later point in time. In order to overcome this problem, Schroth and his colleagues decided to exploit the subsequent photographs taken by those using NAVVIS as a navigational tool. The system has the ability to adapt to changing interior environments.
"From the time that NAVVIS becomes operational in a building, users will take photographs of the environment to localise themselves," he explained. "These images can be used to update the map. Typically, the look of a hallway will not change completely. Visual alterations to posters and door signs, for example, only tend to affect a limited area of a particular location. New photographs enable NAVVIS to update its map whilst simultaneously allowing a user to localise. The system gets better and better over time. You could say that NAVVIS is a crowd-based system."
Whilst Schroth and his colleagues are currently testing NAVVIS around TUM’s main campus, they intend to trial their system inside some of Munich’s museums in 2013. In addition to providing a useful navigational tool, the researchers believe that NAVVIS could act as a platform for augmented reality. I concluded our conversation by asking Schroth how this might work.
"As I said, localisation is currently accurate to about one metre and to between 10 and 20 degrees," he reiterated. "You can imagine that augmented reality might prove very useful within a museum setting. For instance, your current view could be overlaid with additional media relating to an exhibition. The big question concerns the degree to which you want to incorporate augmented reality. If you simply want to receive relevant information about the object at which you are looking, we could achieve that now. However, if you want a system that is perfectly
overlaid with reality – if you are looking for accuracy on the millimetre scale – then extra algorithms would be required to allow your mobile device to precisely track an object. This might be possible in the future but it is not currently part of NAVVIS."
Take the TUM virtual tour and learn more about the visual localisation system by visiting the NAVVIS website