| dc.description.abstract | An automatic surveillance tracking system plays an important role in security applications. In this thesis, we develop a real-time surveillance system for tracking moving objects, like people, animals, vehicles, etc.
Our system consists of three parts. In the first part, we use the background subtraction technique to detect the moving pixels. In the method, we build an adaptive background to deal with the problems of lighting change, and repetitive motions from clutter. In the second part, we remove the shadow and noise in the images to improve the system accuracy. In the third part, we construct the foreground objects with color and shape information. We also use foreground objects’ characteristic to match, track, and predict the position of the moving object.
In the experiments, we consider several different weather conditions such as sunny, cloudy, dusky, rainy hours, and night, and different backgrounds like building, tree leaves, roads, and monitor screens. From the experimental results, we find that the proposed approach can accurately detect and track different moving objects in the different weather conditions, and environments. | en_US |