Indoor Localization Using Camera Phones

Abstract

i Figure 1. Left: User wearing the phone as a pendant, Right: Snapshots of the client running on the phoneThere has been a shift in the focus of indoor localization research from improving accuracy to minimizing infrastructure requirements [4, 6, 1]. The reason is well understood: since location information only serves as a parameter to location-based services, the cost of deploying localization systems should be a minute fraction of the total cost of provisioning location-based services. We demonstrate the possibility of determining user's location indoors based on what the cameraphone "sees". The camera-phone is worn by the user as a pendant( Figure 1) and images are periodically captured and transmitted over GPRS to a web server. The web server has a database of images with their corresponding location. Upon receiving an image, the web server compares it with stored images, and based on the match, estimates user's location. We accomplish this with off-the-shelf image matching algorithms (namely Color Histograms [7], Wavelet Decomposition [2] and Shape Matching [3]) by tailoring them for our purpose. We use three methods for determining location: Naive, Heirarchical and History-based [5]. The advantage of our approach is that neither custom hardware, nor wireless access points are required. Physical objects do not have to be "tagged" and users do not have to carry any device apart from what they already do: a mobile phone. The only cost involved is that of building an image database.We constructed a partial image database for the third floor of the Rutgers Computer Science building with multiple images per "corner" to account for issues such as varying heights of the users, different angles of view, etc. Our experimental results indicate that room-level accuracy can be achieved with more than 90% success probability, and meter-level accuracy can be achieved with more than 80% success probability. The goal of this demo is to illustrate the basic functionality of the system in a limited setting. We will construct an image database for the demo-room on the spot, and demonstrate our system on this database using the Heirarchical approach. We will use a Nokia 6650 phone running a Java client that would capture images and display location, and a server that would host the image database and the location determination algorithms. We will briefly explain the underlying algorithms, the database construction methodology and the key challenges.

Energy-Efficient Boundary Detection for RF-Based Localization Systems

“…First of all, we propose a mobility-aware adaptive sampling enhancement for boundary detection in RF-based localization systems. Similar ideas of trading location fidelity for better energy efficiency have also been proposed in other localization systems [6], [7]. These works will further be discussed in Section 5.…”

“…In [6], the system assumes target moves in constant speed and transmits a new picture to reestimate the location and moving direction when it predicts the target comes close to the corners in the space. The general idea is similar to our adaptive sampling mechanism but our extended scheme can further relax the constant-speed assumption and provide a more practical solution.…”

Seeker-assisted human navigation using smart phones

“…Combining GPS, WiFi and/or GSM, multiple mobile devices can determine their positions [3], [9]. The third one is to propose methods to identify logical locations, as opposed to physical coordinates [10], [11].…”

Probabilistic Inference of Multi-Object Trajectories

“…Furthermore, it requires the location map of all wireless access points which requires either a cooperative infrastructure or wardriving which is expensive and unreliable due to the dynamic nature of Wi-Fi access points. Other authors attempted to leverage other sources of information [4,5] for indoor localization, but all of the existing techniques focus on the localization of a single object. This paper treats the localization problem from a different perspective.…”