Improving car navigation with a vision-based system

H. Kim, K. Choi, I. Lee

Research output: Contribution to journalConference articlepeer-review


The real-time acquisition of the accurate positions is very important for the proper operations of driver assistance systems or autonomous vehicles. Since the current systems mostly depend on a GPS and map-matching technique, they show poor and unreliable performance in blockage and weak areas of GPS signals. In this study, we propose a vision oriented car navigation method based on sensor fusion with a GPS and in-vehicle sensors. We employed a single photo resection process to derive the position and attitude of the camera and thus those of the car. This image georeferencing results are combined with other sensory data under the sensor fusion framework for more accurate estimation of the positions using an extended Kalman filter. The proposed system estimated the positions with an accuracy of 15 m although GPS signals are not available at all during the entire test drive of 15 minutes. The proposed vision based system can be effectively utilized for the low-cost but high-accurate and reliable navigation systems required for intelligent or autonomous vehicles.

Original languageEnglish
Pages (from-to)459-465
Number of pages7
JournalISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Issue number3W5
StatePublished - 19 Aug 2015
EventISPRS Geospatial Week 2015 - La Grande Motte, France
Duration: 28 Sep 20153 Oct 2015


  • Car Navigation
  • Intelligent Vehicle
  • Position and Attitude
  • Single Photo Resection


Dive into the research topics of 'Improving car navigation with a vision-based system'. Together they form a unique fingerprint.

Cite this