Student: Xuan Liu, M.Eng Student
Supervisor: Prof. Mark,Coates
Abstract: Surveillance systems have already become prevalent in many highways and intersections. Visual tracking is an important technique in the automatic real-time surveillance systems; such tracking can consistently locate desired objects in each frame of certain video. This work studies the statistical estimation method -- Particle filter, which is a kind of algorithm to solve Bayesian estimation problems. According to the requirement for vehicle tracking, we choose image differencing and region growing for motion detection and particle filter for tracking. Image differencing is a simple method to detect the moving objects which is adaptive to changes in the dynamic backgrounds. Region growing is to solve the foreground aperture caused by image differencing and variety of vehicles.
[Full Description] [Paper (pdf format)]
|