Research On In-shore Ship Detection From Optical Remote Sensing Image Via Auxiliary Knowledge

With the development of the remote sensing technology,the target detection by remote sensing has attracted more and more attention.The detection of in-shore ship target in the port,which as the important facility,has important significance in port monitoring,military reconnaissance and many fields.However,it is hardly to obtain satisfactory results only depending on the single source of remote sensing image.Especially for the optical image which is vulnerable to the environment,the features extracted from the image are always not enough to meet the requirements of the target detection.In recent years,with the continuous improvement of the knowledge system and information resources,it is feasible to conduct target detection in the help of the auxiliary knowledge.For the task of ship target detection,this paper excavates and summarizes the auxiliary knowledge available and establishes the model.After that,an in-shore ship detection method via the auxiliary knowledge model is proposed.The main work of this paper is as follows:(1)A knowledge model for ship target detection task is established.Firstly,the auxiliary knowledge related is reviewed and method of knowledge expression and modeling is discussed.Then combining the characteristics of ship detection task,the knowledge application pattern and detection process is given.(2)A separation method of the land and sea assisted by the geographic information is proposed.Firstly,the process of segmentation aided by geographic information is determined.Then,the image pre-processing is studied and the method of global segmentation considering neighborhood characteristics is proposed.After that the local segmentation is conducted via excavating and using auxiliary knowledge.Finally the final result can be obtained by combining the local and global result.The method proposed gets good segmentation result by using geographic information appropriately.(3)A ship detection method under the assistance of ship mooring rule is proposed.Firstly,the ship target characteristics in optical images and ship mooring rules are given.On this basis,the process of ship detection is put forward.Then the method to solve the problem brought by adhesion and break of target is proposed.After that the identification method is proposed in the help of ship mooring rules.The detection rate in result shows that the method proposed is efficient.