Back
to indexL. Zhao and C. Thorpe
Summary
Uses LADAR to track cars and predict motion. Lots of equations. Uses kalman filters to detect and predict seperate maneuvers for multiple cars.
Methods
Three parts: car detector, car tracker, and motion estimator. Assume cars are rectangular shapes. Lines in LADAR images are derived from Hough Transform, and then cars are detected from the lines. Motion is classified by constant velocity, constant acceleration, and turning modes. Motion is measured based on the classification. Apply motion results into extended kalman filter, which selects appropriate model for vehicle dynamics. Motion is estimated using EIMM algorithm, which is based on probability and current state.
Keywords
tracking, motion estimation
Rating
7
Bibtex Entry
@misc{ zhao98qualitative,
author = "L. Zhao and C. Thorpe",
title = "Qualitative and quantitative car tracking from a range image sequence",
conference = "International Conference on Computer Vision and Pattern Recognition (CVPR'98)",
pages = "496--501",
year = "1998",
}