Study On The Ship Detection In The Monitoring Video Of Ferry

The thesis derives from the "wisdom maritime platform" project from The Maritime Safety Administration,Zhongshan city,Guangdong Province.This platform requires detection and tracking of the ships,which technically support ship speed monitoring,ship driving violations,oil leak responsibility identification,etc.Based on the actual monitoring data of ferries of Zhongshan city,this thesis focuses on detection and tracking of the ships.General process of the work is as follows:Firstly,thoroughly test three background subtractions including direct subtraction,single Gauss model and mixed Gauss model,then figuring out advantages and disadvantages of them.With direct subtraction,in spite of best real-time performance,information of ships remains on the background and "ghosting" is apt to appear;due to unstable background on water,The Gauss function's variance of the single Gauss model on the background is not converging,thus single Gauss model cannot accurately detect ships;mixed Gauss model has achieved good results in various situations,but due to the large amount of computation,it is difficult to achieve real-time under the current monitoring of the mainstream resolution of the camera.Secondly,improve background updating mechanism of direct subtraction to eliminate the "ghosting".Due to the slow speed of ships,with direct subtraction ships'information remains on the background while updating.Foreground pixel not updating in the improved direct subtraction,so ships' information does not update to background.Thirdly,combine the improved direct subtraction with Meanshift.After testing videos via improved direct subtraction and morphological processing by the designed method,do drift operation on current frame with Meanshift region of previous one,so that "ghosting" remained after ships driven away in picture at the beginning can be eliminated,meanwhile,achieving tracking ships in the picture.Finally,experiment shows that the designed method is able to effectively detect and track ships in monitoring videos of the ferry.