| Summary: | 碩士 === 淡江大學 === 資訊工程學系 === 92 === This paper presents a vision-based traffic monitoring system via analyzing color image sequences of traffic scenes recorded by a stationary camera mounted on a tall building or pedestrian crossing bridge near a traffic light. The system algorithms consist of building initial background, segmenting foreground, removing shadow, locating and tracking vehicles, background updating, detecting velocity of the vehicles, and classifying vehicle type. The YCrCb color space is adopted in our system. It provides us an accurate and robust device for foreground and shadow detection. Our system can accurately locate vehicles, track vehicles, detect velocity of the vehicles, and classify vehicle type from image sequences. The system can be applied for detecting traffic violations and monitoring the traffic condition. It is fast enough to operate in real time, insensitive to lighting and weather conditions, and only requires a minimal amount of initialization. Experimental results from town scenes are provided which demonstrate the effectiveness of our system.
|
|---|
