Affiliation of Author(s):电子信息工程学院

Journal:Lect. Notes Eng. Comput. Sci.

Abstract:To improve the accuracy of SAR sea ice image classification, a new semi-supervised sea ice segmentation is proposed. Different initial label field and feature field of MRF energy function have important impacts on the final segmentation results. The initial label field of energy function is semi-supervised by sea ice recognition result. And the feature field is improved by extracting GLCM feature,The optimal energy function of the MRF model is obtained by simulated annealing method. The weighting parameter of the feature field is taken as a function of the annealing temperature. The influence of feature field on the classification result will auto-adjust to the annealing temperature. Finally, the three classes result of sea ice is accurately and boundaries distinct. Experiments demonstrate that the proposed algorithm is able to successfully segment various SAR sea ice images and achieve improvement over existing published methods including the SA MRF, GMM, and K-means clustering.

ISSN No.:2078-0958

Translation or Not:no

Date of Publication:2017-01-01

Co-author:Fei, Xuanjia,gf,Xing, Shiyu,Leung, Henry

Correspondence Author:孔莹莹,Fei, Xuanjia,Katherine Kong