| PolSAR system obtains the echo information according to the combination of the polarization state of the transmitter and th e receiver.It works in the microwave section and has the characteristics of full day,all-weather and strong penetration capabilities to provide abundant ground information.Compared with other remote sensing data,it has a greater advantage.The near shore ship detection technology based on PolSAR image s has a broad development prospect in port monitoring,ship rescue and coastal dock planning.It is also unaffected in extreme climate conditions and complex scenes.The detection results are deep in the development of the coastal cities,the management of transportation and the safety of the sea area and it is also a hot spot of research and application at present.This paper focuses on the nearshore ship detection methods,mainly as follows:Firstly,the theoretical basis of the PolSAR system,the representation of polarimetric data and the methods of feature extraction are introduced,including the extraction of polarization features and texture features,sea-land characteristics and ship characteristics are also analyzed,in order to make the foundation for the follow-up of the ship target detection.Secondly,a sea-land segmentation method based on connected domain processing and scale change is studied.Based on the difference between land and sea,using mathematical morphology and Freeman chain code to adjust the connected domain on land to improve the segmentation results,the fine sea land partition is obtained by using the seed filling algorithm.In the large scenario,a fast land and sea segmentation algorithm is constructed by scale transformation,which is used to improve the speed of sea-land segmentation.The two value image after sea-land segmentation is applied to land mask,in order to reduce the interference of land objects to subsequent ship detection.Thirdly,an anomaly detection method based on contrast enhancement and a supervised detection method based on sample data are used to detect offshore ships.Considering the difficulty of obtaining prior information,the former method does not rely on the samples,only uses polarizing whitening filter to enhance the ship sea contrast,then uses the romberg integral algorithm to obtain the nominal factor and uses window moving method to determine the pixels;the latter method relies on a small number of samples to obtain the best weight vector and the best Lagrange multiplication of the support vector machine,the whole image is traversed to detect ships based on the decomposition surface equation and discriminant formula.They make a good transition for the detection of docked ships.Finally,the docked ships are closely related to the docks,the size of the ships and the ways of stopping are not the same,the docked ships are divided into the land area in the sea-land segmentation,the environment is relatively complex,and the land interference is much,so the docked ship detection becomes very difficult.In view of this problem,the land is firstly masked by two value image of sea-land segmentation which is propelled the width of the ship or ship cluster along the coastline,then a detection method of docked ships based on computational intelligence algorithm to adjust SVM parameters is constructed to reduce the false alarm rate of the docked ship detection. |
PDF Full Text Request