Research Of Inshore Ship Detection Based On High-resolution Remote Sensing Images

With the rapid development of optical remote sensing satellites,ship detection and identification based on large-scale remote sensing images has become a significant maritime research topic.Because of a wide range of applications,it has attracted the attention of the military and civil field.Compared with traditional ocean-going vessel detection,inshore ship detection has received increasing attention in harbor dynamic surveillance and maritime management.However,because the harbor environment is complex,gray information and texture features between docked ships and their connected dock regions are indistinguishable,most of the popular detection methods are limited by their calculation efficiency and detection accuracy.In this paper,we focus on the theoretical research of the important steps of inshore ship detection that include rapid water-land segmentation,significant berth extraction and candidate target identification.And the main works are as follows.First,a rapid water-land segmentation algorithm with adaptive ability is proposed in this paper.Based on the peak characteristics of gray feature distribution and texture neighborhood variance distribution,an adaptive segmentation threshold could be obtained to preliminarily extract the harbor water areas.According to a series of morphological methods,false candidate regions that are obviously incorrect can be deleted.This method can automatically obtain the threshold and complete the rapid water-land segmentation without GIS information in diversified port areas.Second,a significant berth extraction method based on spatial adjacency feature is proposed in this paper.A docked ship presents a protruding characteristic in a smooth area along the coast.Considering this characteristic,intersected two-dimension scanning(ITDS)is used to extract the candidate.To overcome the various locations of docked ships and their different location angles,an omnidirectional intersected two-dimension scanning(OITDS)technique is designed.This method could scan as many suspected protruding bulges as possible in different directions.Third,a decision mixture model(DMM)is proposed to identify real ships from candidate objects.To increase the accuracy of candidate region identification,the key part sub-model,a whole ship sub-model and a context sub-model of ship are integrated into the proposed DMM.This comprehensive decision method fully considers the adaptability of target deformation and increases the robustness on unsatisfactory scenes.On the basis of the above three studies,a hierarchical framework for inshore ship detection in large-scale harbor remote sensing images is presented in the end of this paper.This method aims at rapidly and accurately getting the inshore ship from interested port by using a wide range of remote sensing images.Experiments on a large number of large-scale harbor remote sensing images verify that the proposed method is effective and robust when applied to unsatisfactory scenes.Compared with typical methods,the proposed method also achieves better detection results.