A VEHICLE-CANDIDATES VIDEO DETECTION METHOD BASED ON SEGMENTATION OF COLOR ROAD AREA

Chen Zhaoxue,Shi Pengfei
DOI: https://doi.org/10.3969/j.issn.1000-386X.2007.10.011
2007-01-01
Abstract:A vehicle-candidates video detection method based on segmentation of color road area is proposed.The road area is segmented at first from a single colorful video frame based on the feature of road area colors distribution.Either dynamic or static vehicle-candidates could be detected by background difference method and the related post-process course after the background initialization of road area.The background is trained and updated simultaneously just at the moment of vehicle detection by adoption of a double-background strategy,which overcomes the problems of error accumulation and sensitivity limitation over light conditions that occur during background updating of traditional methods.With simple and robust implementation,this method could satisfy the real-time requirements of background updating and vehicle detecting in traffic video surveillance system.Its validity is proved by the experimental result.
What problem does this paper attempt to address?