Research Project P13:
Where Am I?

Research Project Goal

You are given a large database of sparsely sampled city street views tagged by GPS location. Given a novel query view compute one or two closest views from the database and the GPS coordinates of the new query view.

Sample Data

Dataset of city street views tagged by GPS location research.microsoft.com/iccv2005/Contest/Data/Final5.html More details about the units of measurements and how to compute longitude and latitude can be found at research.microsoft.com/iccv2005/Contest. For the calibration information contact the instructor.

Tasks

• extract suitable features (points, lines, SIFT fetaures)
• implement image matching by finding the closest view from the database (Simple voting might be your first choice, but more sophisticated techniques are applicable here.)
• once the two closest views have been determined, compute the relative motion between the query view and closest views using the epipolar geometry and RANSAC and use triangulation to compute the GPS coordinates of the query view

Challenges and Open Problems

• image representation, which facilitates efficient matching
• robust computation of the epipolar geometry
• How would you tackle the problem of indexing the database if it were comprised of the millions images of the whole city ?

Research Project Status

student names here

Point of Contact

Jana Kosecka

Midterm Report

not yet submitted

Final Report

not yet submitted