Visual Surveillance System Of Invading Pirate

Pirates often haunt when ships travels in coastal water, especially in the area of Somalia. That largely poses a great threat to the international maritime security. Therefore, this thesis tries to design a visual surveillance system that identifies pirates, which lets the crew to find pirates in the first time and to take appropriate measures to deal with pirates.Firstly, in this thesis, some of the existing mainstream motion detection algorithms are summarized and analysed, and some knowledge of images is added. Visual surveillance system designed in this thesis uses a method of codebook background modeling which can clear the sea wave caused minor noise.Secondly, codebook modeling has a good result on a regular interference, but it is not good at random interference. So on the basis of codebook modeling we use a connected domain method to remove random noise.Then, according to separating out the foreground objects, we use detection and tracking of target. And this thesis adopts the method of motion template. Through the Direction of moving objects are predicted, we determine its direction of movement. If this direction in the period of time points to the ship, it is a pirate ship, then gives an alarm and alert in time.Finally, through the independent threads manage of each link, this thesis uses drectshow toolbox to set up three Filter Graph Manager that collects video frame in each link. Before the main thread calls the handler function for processing, we will transform Drectshow frame format into image storage format of OpenCV and call processing algorithm to analyze it. This completes the process of building the system. In this paper, experiments show that the system is effective.