3D personal navigation in smart phone using geocoded images

Abstract

The research of three-dimensional personal navigation in smart phone is more and more popular nowadays. Such kind of 3D navigation applications are mostly based on 3D model constructed in the computer, then importing the model to the mobile phone. Comparing with computer, the mobile devices typically have the bottlenecks of weaker CPU, less memory, lower bus bandwidth and limited power supply. The thesis presents a new solution to mitigate these technical restrictions of 3D model visualization in smart phone and achieve a better performance. The thesis is based on the project of "3D Personal Navigation and Location-Based Service for World Exposition 2010" funded by TEKES, replacing 3D visualization model engine with geocoded images searching engine in smart phone. The research delivers a 3D personal navigation software running in a Nokia 6710 handset.

The major difference of the proposed solution with other similar 3D mobile navigation solutions and the previous 3D PN application is the visualization engine, which is based on geocoded images, rather than the developed projects on the basis of 3D model.

The research generates a topological network and landmark database, and geocodes images snatching from 3D model which is constructed on computer. The system implements the functions of navigation in querying, routing and map matching, and develops the visualization search-engine for the geocoded images. To honestly evaluate the performance and improvement of the proposed solution several tests are carried out in the area of Tapiola, Espoo, Finland, including a field test to evaluate the positioning accuracy and the real-time performance, and a consumption test to prove the low cost of system.

The proposed solution costs lower power consumption, CPU and memory, in higher refreshing rate,and have better visualization performance as well. In addition, this solution provides an optional way to decrease labour and working time to build up a similar solution by using real photos instead of complex 3D model construction.