基于改进谱聚类的合成孔径雷达溢油图像分割算法

Segmentation algorithm of SAR oil spill image based on improved spectral clustering

为了解决传统谱聚类算法对大尺寸海洋图像难以进行有效计算的问题,提出一种改进的谱聚类算法.采用分块方法将原始图像分割成多个子图,同时结合随机采样算法利用采集的样本估计全局样本,在保证分割精度基础上大大降低计算复杂度,有效地处理高维图像.针对随机采样的不稳定性,采用多次采样聚类并结合大多数投票的方法,得出最终的分割结果.仿真结果显示,改进算法可以有效降低计算复杂度,并保证聚类算法计算复杂度的减少与图像大小成正比,分块方法和多次聚类结果的融合可以大大提高溢油目标分割的精度.

In order to solve the problem that classical spectrum clustering algorithm can not calculate effectively for the large sea images,an improved spectral clustering algorithm is proposed.The partition method is used to divide the original image into multiple sub-images,and the collected samples are employed to estimate global samples combining with the random sampling algorithm.Thus,the computational complexity is greatly reduced on the basis of guaranteeing segmentation accuracy,and high dimensional images are processed effectively.Multiple sampling clustering combining with the majority voting method is used to obtain the final segmentation results in view of the instability of the random sampling.The simulation results show that the improved algorithm can reduce the computational complexity efficiently,and guarantee the decrease of computational complexity proportional to the size of images.Meanwhile,the hybrid of partitron method and multiple clustering can make the segmentation of oil spill target reach high precision.