Coastline Detection Algorithms Based On Space Domain Statistics For SAR Images

Coastline detection of Synthetic Aperture Radar(SAR)images has significant meanings in the fields such as maritime management and automatic navigation.It's a great challenge in coastline detection of SAR images,because of the complexity of land and values and bright reflections at sea areas.In terms of space domain statistics,edge detection and active contour model,we make studies on coastline detection of SAR images with bright reflections in sea areas in this paper.The main work of this paper as follows:(1)A modified coastline detection algorithm based on the space domain second-order statistic is presented.Most existing algorithms are difficult to detect coastlines of SAR images with bright reflections at sea areas.To overcome such a problem,the space domain second-order statistic is presented to suppress the brightness of sea.The region between land and sea is obtained by the statistic,then the edge detection algorithm based on Gaussian-Gamma shaped bi-windows is applied to detect the coastline.Experiment results show that the modified algorithm can achieve good performance of coastline detection in SAR images with bright reflections in sea.(2)A modified coastline detection algorithm based on the directional statistic with variance of means and the improved regularization term of Gamma distribution ACM is presented for coastline detection.Due to realise the coastline detection of some SAR images with bright reflections at sea areas,a directional statistic with variance of means is presented to suppress the brightness of sea,firstly.Secondly we can detect coastline in the rough region of land obtained by the statistic.Then a improved regularization term based on diffusion coefficient of conditional information is presented.Subsequently,a detailed theoretical derivation of the modified model and level set evolution equation are provided.The algorithem has good performance for some SAR images with bright reflections at sea areas that the local homogeneity exists.(3)An improved local window Gamma distribution fast level set coastline detection algorithm is provided.Firstly rough land area is acquired by suppressing the brightness of sea according to the directional statistic with variance of means,and the area is used as initial contour.To overcome the problem that some active contour models need long time to achieve coastline detection.The improved contour evoluting speed term of local window by Gamma distribution is presented and is used to update the level set with a quick speed.Because some small areas occured in the complicated areas in land,region judgement is provided to update the level set of the small regions.Finally the coastline can be detected.The effectienty of the modified algorithm is validated with experiments..