Research On Detection Algorithm Of Moving Ship In Inland River Video Surveillance

Video supervision technique is widely used in the management of inland shipping,not only can monitor the state of ship shipping,also can achieve the functions of ship tracking and positioning,information query,emergency processing and other functions.It has become a new direction for the future development of shipping management.In the actual inland river scene,it is still difficult to accurately extract the ship's target because of the interference of the change of closing line,the fluctuation of the water surface and reflection of its own reflection.This paper starts with the human visual attention mechanism,in the video series to study the motion of ship detection algorithm with the experiment.The main work and research results are summarized as follows.Firstly,this paper research on common algorithms of moving target detection,determining the background difference method as the target detection algorithm most suitable for the water environment,aiming at a large number of water ripple region existing in the foreground,proposed to weaken the wave interference according to the contrast of Saliency.Secondly,in view of the MSS algorithm is to obtain salient region of saliency is low,easy to cause saliency target incomplete problem has made the improvement.To solve this problem,this paper combines gauss pyramid to carry out images multi-scale sampling,and then in the visual feature extraction section introduces edge features of an image,to maintain the integrality of the saliency information,and then to enhance the saliency degree of the saliency region through different scale of saliency map fusion.The comparison experiment with the existing algorithms shows that the proposed algorithm can better highlight the saliency region and get the internal uniform saliency objects.Finally,the improved MSS algorithm is applied to ship target detection,which can not only suppress the water wave,but also locate the target area of the ship.But the salient region includes the ship targets and other stationary salient region.Therefore,we propose to combine the codebook model background detection algorithm with the improved MSS algorithm to complement each other and complete the ship target detection in the surveillance video.Experiments show that this algorithm can effectively extract the ship target,its experimental effect is better than other ship detection algorithms.