基于改进差异算子和Gabor_ELM的SAR图像变化检测算法

SAR IMAGE CHANGE DETECTION ALGORITHM BASED ON IMPROVED DIFFERENCE OPERATOR AND GABOR_ELM

针对差异图构造单一、抗噪性差及纹理特征不足等问题,提出一种基于改进差异算子和Gabor_ELM的SAR图像变化检测算法。该方法结合改进对数比的高效性和邻域比的鲁棒性,提出一种改进差异算子,再通过对差异图提取Gabor纹理信息结合层次模糊聚类和极限学习机(ELM)来实现Gabor_ELM变化检测算法。实验表明,与FLICM、MRFFCM、ELM及PCANet算法相比,提出的算法能较好地保留变化细节信息且具有较高检测精度。实验数据正确率达到98%以上,Ottawa数据的Kappa系数达到0.94。

Aimed at the problems of single construction,poor noise resistance and insufficient texture features of difference map,a SAR image change detection algorithm based on improved difference operator and Gabor_ELM is proposed.Combining the efficiency of the improved logarithmic ratio and the robustness of the neighborhood ratio,the method proposed an improved difference operator,and extracted Gabor texture information through the difference map combined with hierarchical fuzzy clustering and extreme learning machine(ELM)to achieve Gabor_ELM change detection algorithm.Experiments show that compared with FLICM,MRFFCM,ELM and PCANet algorithms,the proposed algorithm can better retain the change details and has higher detection accuracy.The correct rate of the experimental data is over 98%,and the Kappa coefficient of the Ottawa data reaches 0.94.