Inland Moving Ship Detection Based On Background Subtraction

Inland waterway transportation plays a crucial role in China's transportation industry.However,the diversity of ships and the complexity of the waterway environment have brought difficulties to the maritime authorities.In order to strengthen the safe supervision,inland maritime supervision system is gradually building,where electronic cruise system as the primary and advanced means.However,at present,electronic cruise system takes the main way to manually monitor videos,costing a lot of manpower and resources.With the rapid development of image processing technology,video processing will be applied to electronic cruise system,detecting moving ships automatically is benefit to promote its intelligent.At the same time,inland moving ship detection as low-level visual task,provides the basis and premise for advanced visual tasks,for example,ship tracking and behavior analysis of ship.Therefore,it's great practical significance to carry out the research on moving ship detection.The contents of this paper are summarized as follows:(1)The differences and relations between inland moving ship detection and object detection,moving object detection,moving vehicle detection,maritime moving ship detection are conduct.The key and difficulties of inland moving ship detection provides direction for algorithms.Evaluation criteria provides objective standard for quantitative analysis.(2)A novel inland moving ship detection based on sparse representation and saliency detection is proposed.Firstly,background model is established by random sampling and sparse representation.Then,the significance region in inland image is extracted by saliency detection.The bitwise logical AND operation is performed between difference image and saliency map to obtain the final result.The background update mechanism is put forward to adapt the change of environment.The method attacks the challenge when ships suffers serious cavities due to their large size,relatively low speed,and uniform color.(3)The optimization strategy for self-balanced sensitivity segmentation is proposed.Firstly,the validity for dynamic background and shadow is analyzed through introducing the principle of self-balanced sensitivity segmentation.Connected domain processing is adopted to solve the problem of missed detection for small ship at the edge of image.Multi-threaded programming is used to decrease time cost.Both qualitative and quantitative evaluations on several challenging inland videos demonstrate that the method outperforms several state-of-the-art algorithms in terms of efficiency and accuracy.(4)A comprehensive,representative and public inland waterway moving ship detection benchmark is developed with data collected from local maritime bureau.Inland moving ship detection algorithm platform is designed based on MFC and OpenCV,which integrated a variety of algorithms and has high efficiency and stability,providing a tool for scientific experiments for peers.