Algorithm And Implementation Of Vessel Size In Video Detection Based On Moving Trajectory

Because of the fluctuations in water and irregularities of ships' size,ships may hit bridges and strand.How to detect ships' size through video surveillance does matters on the security of inland waterway.However,it brought many difficulties to detect and track the moving ships,because of the complexity of the water scenes and ships blocking each other,which affected the accuracy of the size detection.This paper is originated from "Ship Networking" national major sub-projects,study the difficulties and solutions of the motion vessel based on the video image detection,tracking the river,and the use of image parameters to calculate the size of the ship.Background subtraction is the commonly used approach in Ship detection phase,but the detection of moving ships are not complete,which may cause difficulties to extract and split the ships.For some practical problems like segmentation and near ships merging as one,the paper improves ship segmentation method.On tracking of moving vessels,Kalman filter tracking is a commonly used method.For Kalman's multi-target tracking can't solve the problem of occlusion,the paper introduces the histogram tracking and improves with vector polymerization approach.Compare with histogram tracking,the method is applicable in the light change scene,which can used as supplementary tracking method when the ship is shadowed.When measuring the length of the vessels,firstly for the irregular shape of the vessels,the author uses a rectangular frame to approach ship,and converted to calculate the length and width of the rectangular frame.For the disadvantages of rectangular box,the paper makes an intersection between linear path and contour.It used the intersection as ship shaft length and endpoint width.Compared with rectangle method,this method ensures length and width endpoint are both on the edge of the ship.Finally,experimental results demonstrate the effectiveness of the algorithm.The main innovations of this paper are:? Improves extraction methods of complete moving vessel;merge area among moving vessels,and avoid the problem of blending ships nearby;? Put forward to coherence vector tracing method which made up Kalman tracking's failures in occlusion;overcome the problem that histogram is difficult to distinguish between different target and track down a complete track of occluded ships;? Optimize calculation method of ships' image size.Changing from only measuring the width and length of rectangle to tracking through the intersection point of the straight line and contour to calculate the length and width of ships.It avoids the disadvantages of the rectangle vertex is not on the boat along,so can get more accurate size of the ship.