Research On Detection And Tracking Method For Shipping Vessel

Maritime transport has long been a serious threat to maritime security issues represented by piracy crimes.In response to the problem,the thesis combines the merchant ship intelligent visual water cannon to carry out related research on shipping vessel monitoring technology.In view of the ship detection problem,the thesis proposes a ship monitoring algorithm which combined with object detection and tracking method.The method combines the Faster R-CNN algorithm and the image haze removal algorithm to ensure that the detection model has good stability and effectiveness against severe weather such as rain and fog.The experimental results show that the ship detection algorithm has good effects under rain and fog conditions.Then,based on the suspicious ship position information obtained by Faster R-CNN algorithm,the appearance model is established,the stable tracking of the ship is realized by the relationship between the appearance model and the moving state of the object between adjacent frames in the video sequence,the tracking experiment results show that the KCF can ensure the tracking precision while tracking the ship in real-time.The thesis presents a ship detection method based on data augmentation and Hard Negative Mining,aiming at improving the precision of ship detection model.The method combines the idea of hard negative mining,and proposes a bootstrap sampling method based on hard negative mining,strengthening Faster R-CNN algorithm to study difficult samples in dataset.At the same time,combined with cutting,flipping and adding noise,to achieve equalization of sample sizes of various types of ships.The experimental results show that the method can significantly improve the precision of the ship detection model.