Segmentation algorithm of fuzzy clustering on side scan sonar image

Wang Lei,Ye Xiufen,Wang Tian

DOI: https://doi.org/10.13245/j.hust.2012.09.011

2012-01-01

Abstract:Aiming at the characteristics of background areas in the sonar imagery,the side scan sonar image was decomposed into several intrinsic mode functions(IMFs) and a residue with the bidimensional empirical mode decomposition(BEMD).Features of the targets and shadows were enhanced and the noise was reduced.In order to use the spatial relationships between pixels,Gauss Markov random field(GMRF) was used to extract the local texture,and the mis-segmentation was reduced.Finally,a new fuzzy clustering algorithm was proposed,and its clustering criterion and distance function were rewritten.By the comparison with other traditional clustering methods in sonar image segmentation field,the proposed algorithm was proved to be superior to others.