Multiple Vehicle Detection and Tracking in Hard Real Time

Margrit Betke and Esin Haritaoglu and Larry S. Davis

Back to index

Summary

Older but widely cited paper. Use relative motion between car with camera and its environment to detect potential cars.Narrowed fown by objective function searching for typical features. Lots of emphasis on real-time OS, Maruti.

Methods

3 subsystems for vision: car detector, process coordinator, and tracker. Car recognition breaks into passing and distant cars. Passing cars are detected by relative motion. Distant cars are matched against templates. Process coordinator forks a process to track a car (<<-- Interesting approach). Cars become "stable" if they are tracked with high scores over several frames.

Keywords

real-time, feature detection, template matching

Rating

7

Bibtex Entry

@techreport{ betke96multiple,

author = "Margrit Betke and Esin Haritaoglu and Larry S. Davis",

title = "Multiple Vehicle Detection and Tracking in Hard Real Time",

number = "CS-TR-3667",

year = "1996",

url = "citeseer.ist.psu.edu/betke96multiple.html"

}

Back to index