Ego-localization by Matching In-vehicle Camera Images with Aerial Image

In recent years, the ITS technologies are researched actively. Especially, an accurate ego-localization estimation is becoming very important function in this field. By obtaining vehicle position accurately, it is possible to realize smart driving support system when changing driving lanes or entering cross-road. This paper proposes accurate ego-localization method consisting of two steps: (i) projection of in-vehicle camera images onto bird’s-eye view, and (ii) sequential image matching of projection images and an aerial image. An aerial image is used for obtaining global position of a vehicle without accumulation error. We evaluated our method by comparing estimated vehicle position with gold standards. Experimental results showed that the average estimation error was 0.61m (SD: 0.27m).

[1]  Paul A. Viola,et al.  Alignment by Maximization of Mutual Information , 1997, International Journal of Computer Vision.

[2]  Gérard G. Medioni,et al.  Map-Enhanced UAV Image Sequence Registration , 2007, 2007 IEEE Workshop on Applications of Computer Vision (WACV '07).

[3]  Robert C. Bolles,et al.  Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography , 1981, CACM.

[4]  Takeo Kanade,et al.  An Iterative Image Registration Technique with an Application to Stereo Vision , 1981, IJCAI.

[5]  Alexander Bachmann,et al.  Visual features for vehicle localization and ego-motion estimation , 2009, 2009 IEEE Intelligent Vehicles Symposium.

[6]  Takeo Kanade,et al.  Transforming camera geometry to a virtual downward-looking camera: robust ego-motion estimation and ground-layer detection , 2003, 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings..