handheld augmented reality

Augmented Reality Anywhere and Anytime   

Projects

   Social Augmented Reality

   Information Presentation

   Real-Time Self Localization

   Structural Modelling

   AR Graphics

   AR Navigation

   Augmented Reality Games

   Past Projects

 

Technology

   Hybrid SLAM

   Panorama SLAM

   Planar SLAM

   Model-Based Tracking

   Marker-Based Tracking

 

Software Libraries

   Studierstube ES

   Studierstube Tracker

   Muddleware

   ARToolkitPlus

 

More

   Videos

   Media/Press

   Team

   Publications

   Collaborations

   Student projects

   FAQ

   Site map

 

The Handheld Augmented Reality
Project is supported by the
following institutions:

Qualcomm

 

Christian Doppler Forschungsgesellschaft

 

Graz University of Technology

 

 


 

Come to ISMAR 2011

ISMAR2011

 

Follow us on twitter

Map Tracking

Overview

In 2007, Rohs et al. created a software for Symbian phones that tracks maps, which are outfitted with regular grids of dots, and tracked with 2.5 DOF [link]. Our map tracker is similar to this work but performs six degree of freedom pose estimation from almost any map or other uniquely textured planar object. Maps have to be prepared by adding small black dots that serve as reference point for tracking. These dots are small and cover only about 1% of the map's area. Hence the map remains mostly unmodified.

The map tracker has been implemented by adding two new sub modules to our Studierstube Tracker. Hence it can make use of Studierstube Tracker's advanced features such as fast feature detection and pose estimation. It runs on Windows Mobile and Symbian at frame rates of 15-30 fps for a complete application, depending on the phone and the application itself.

Although maps are planar and mostly viewed from afront, the map tracker treats this as a full six degree of freedom problem. It therefore allows rotating the map arbitrarily and tilting the phone to ~45°. This makes the map tracker suitable to track maps or other planar surfaces from almost any position. E.g. it is highly suited to be used for AR board (table top) games.

The following video show the map tracker in action running on an unmodified iPAQ smartphone. The map has a size of ~70x70cm and is made up of 9x9 dots (forming 8x8=64 cells to track). Tracking works as long as a at least a single cell is visible. The over application performance in this video is about 30fps. While the tracker ifself runs faster, the overal performance is limited by the framerate of the camera, which is 30fps.

 

Map tracking on an iPAQ smartphone. A high quality version
of the video can be downloaded from our media section.

 

The following video shows the same application (ported to Symbian) running on an Nokia N95 smartphone.

Map tracking on a Nokia N95 phone. A high quality version
of the video can be downloaded from our media section.

 

How does it work?

The map tracker has benn implemented using the regular Studierstube Tracker pipeline. Two new modules (called "features") have been added: a circle detector and a grid detector. At run-time the following steps are performed for every frame:

  • Thresholding
    The map tracker thresholds the image and automatically extracts dark areas.

  • Contour following
    The contour follower searches for connected regions.

  • Circle detection
    The circle detector checks all extracted contours (connected regions) using a simple ellipse fitting algorithm.

  • Grid detection
    All circles are then handed over to the grid detector that tries to reconstruct the regular grid of the dots on the map.
    The grid detector then extracts the patches between the dots at a resolution of 65x65 pixels.
    This high resolution for the patches is required in order to track maps with high frequencies (small structures such as buildings or streets at a size of a few millimeters), which would otherwise create aliasing and hence result in bad detection quality.

  • Template matching
    The template matcher uses a Gaussian downsampling to reduce the patch size to 32x32 pixels and compares it to all patches in the database.

  • Pose estimation
    Correctly detected cells are used for estimating the camera's pose relative to the map resulting in a 6DOF pose tracking.

 

IPCity

The map tracker has been developed as part of the IPCity project.

 

 

copyright (c) 2014 Graz University of Technology