Multiple Ship Tracking In Inland Waterway

Inland waterway transportation plays a crucial role in China’s transportation industry. However, because of the complexity of the types of ship and the navigational environment, the load of maritime administration in inland waterway is growing. Moreover, there exist some segments that are narrow, curve or dangerous. Those all make the administration costly. Recently, along with the development of the information technology, intelligent video technology is fast and extensively applied into the maritime administration system. Thereinto the CCTV system acts as an eye in the electronic cruise. It monitors the maritime all-weather and all-time. Unfortunately, because of the limit of the intelligent level, it still dependents on artificial participating so as to obtain useful information from those monitoring videos. So, it is significant to develop the ship tracking algorithm in inland CCTV video.In this thesis, we research and develop a multiple ship tracking algorithm in inland CCTV video. My main research content is as follows.Firstly, we summarize the types of ships in the inland waterway surveillance. Then we utilize the histogram of oriented gratitude to describe all types. Moreover, according to the characteristic of inland waterway, we also make some analysis to the surveillance videos.Secondly, we obtain the appearance models of interesting ships by HOG combined with fuzzy SVM. Then we complete the multiple ship tracking in inland waterway based on the deformable part model. We consider into the spatial relationship between ships. Precisely because of the spatial relationship, we can avoid the influence of the similar ship so that to keep the accurate tracking.Lastly, when occlusion occurs between multiple interesting ships, most useful information of occluded ships will be lost. Especially exiting some similar ships, our tracking algorithm may fail to track. In order to solve the problem, we propose to utilize the least square method to fit the ship’s trajectory. When occlusion occurs between ships, we use least square method to tracking. Or we use above-mentioned multiple ship tracking method.A large number of experiments indicate that our multiple ship tracking algorithm in inland CCTV video can complete accurate multiple ship tracking. And our resolution for occlusion is an effective method. They can also complete accurate and robust ship tracking for some main types of ships in inland waterway.