Ship Extraction In High Resolution Optical Remote Sensing Data

The 21st century is ocean era, vessel is the common transportation for ocean de-velopment, and vessel monitoring is an important task for coastal states to protect na-tional maritime rights and interests and to maintain maritime security. Presently, Re-mote Sensing (RS) technology develops rapidly, image is acquired easier and resolu-tion and quality increase, so RS can be widely applied in civil and military. Duo to the superiority of RS, the vessel extraction from RS image is of great significance and becomes an important research direction of target detection.This paper researches ship extraction from high resolution images(HRI).Firstly, through referring relevant literature, understanding key technology and present situa-tion of the ship extraction, the paper designs an object-oriented extraction algorithm. Secondly, introduce the object-oriented workflow, common methods and key tech-nologies. For object-oriented vessel extraction, the paper employs multi-scale seg-mentation. This paper focuses on the principles of multi-scale segmentation, then analyses the relevant parameters, summarizes the guidance parameters, applies the maximum area method to determine the best segmentation scale, eventually gains good results. After that, in the case of multidimensional characteristics of image ob-jects, the paper employs fuzzy classification, deeply research the principle and im-portant three steps, analysis the features of the vessel, summarizes the features set and uses the proper membership function, and achieves finally the optimum features set. And experimental results show the extraction algorithm can accurately extract the ship at sea, and the extraction accuracy is 88.1% and has good stability.Object-oriented vessel extraction method efficiently extracts the ship of HRI. Using HRI for vessels extraction provides a new technique for vessels monitoring and maritime rights and interests protecting, which has important significance.