What’s That? : A Location Based Service Department of Computer Science and Engineering University of Minnesota Presented by: Don Eagan Chintan Patel
Agenda Motivation Problem Statement Related Work Challenges Contributions Implementation Validation Conclusion Future Work Question(s)
Motivation Rapid, Sustained Adoption of Smart Phones 50+ million iPhones Smart Phones Market Growing at ~30% Under Utilized Hardware Features Compass, accelerometer etc. User Friendly Queries “What’s that?” vs. “What’s around location XYZ?”
Motivation How good is that?What’s out there? What’s behind that? What’s that?
Problem Statement Input Location and directional information provided by a mobile device Output An easy to use tool to identify entities in unfamiliar settings User provided with relevant data. identification, reviews, menu, hours, etc. Assumptions / Constraints Availability – The application must execute on commercially released operating system(s) and hardware. Affordability Data set limited to a small number of rectangular objects Pitch of mobile device not taken into account
Related Work Compass/ Directions is Used for Gaming Augmented Reality Applications Layar: GPS + Camera GeoVector: Similar Usecase Layar combines GPS, camera, and compass to identify your surroundings
Challenges Efficient Object Identification Avoid Linear Scan
Contribution Novel Object Querying Approach Combining location + direction Better than Linear Solution
Solution Application built on Android platform Open Source Decent Documentation Developer Tools Small but Real-world Dataset University’s Mall area Application Partitioned into Layers Flexible Architecture
High-level Architecture Data Access Data Processing UI Location & Direction Results Request for ObjectCandidates
Demo
Data Access Layer Implementation Integration of Java open source code Initialization CSV file Entity locations and attributes saved to QuadTree Query Processing Bounding box built from point location and direction Linear search performed on QuadTree results Line intersect processing used to determine entities Entity list returned on all objects falling on directional line
Entity Selection Johnston Walter Smith Morrill Tate Vincent/ Murphy Kolthoff Northrop Ford 400 meters 61 meters
Line Rectangle Intersection Johnston Walter Smith Morrill Tate Vincent/ Murphy Kolthoff Northrop Ford
Functional Validation Johnston Walter Smith Morrill Tate Vincent/ Murphy Kolthoff Northrop Ford 7° 16° 26° 85° 99° 156° 164° 200° 210° 261° 276° 328° 339° 348°0°
Functional Validation Description Distance from Walter (feet) Distance from Walter (meters) Calculated Range of Identification Actual Range of Identification Inside Walter Library-0-90 West wall center Outside Walter Library km boundary * Walter 20° 10° 30° 40° 50° 60°
Accuracy Testing Walter 5729ft ft. 1°5° Description Distance from Walter (feet) Distance from Walter (meters) Test Degrees Test Degrees Resulting in Identification 100 ft. inc for every degree ,89,88 90, ft. inc every 5 degrees ,85,84 86,85
Conclusion Implemented a spatio-directional system that allows querying visible but unknown objects by consuming location and directional sensor information provided by a mobile device.
Future Work Handling Irregular Shapes (non-rectangular) Larger Dataset Inclusion of Pitch – More Closer to the Real-World Hardware Testing
Questions ? ? ?