改进的区间二型模糊聚类遥感影像变化检测

Remote sensing image change detection based on improved interval type-2 fuzzy clustering

遥感影像的复杂模糊性问题会干扰影像变化检测的结果,可引入区间二型模糊C均值聚类算法解决此问题,但算法参数的随机性导致检测结果不稳定。本文首先利用局部最优解优化萤火虫算法中的候选解,引入可变步长因子,以此自适应寻优区间二型模糊C均值聚类算法的模糊因子;然后结合寻优得到的模糊因子进行区间二型模糊C均值聚类,迭代更新隶属区间提取变化信息;最后通过基于复合梯形法则的加权Karnik-mendel算法降型和解模糊优化聚类中心,依据最大隶属度原则判断变化类型。通过试验验证,本文方法得到更优模糊因子和更精确的聚类中心,具有更好的稳健性,变化检测精度得到提高,检测得到的变化区域更加精细。

The complex fuzziness of remote sensing image can interfere with the result of image change detection so that the interval binary fuzzy C-means clustering algorithm is introduced to solve the problem.But the randomness of algorithm parameters will affect the accuracy of change detection.In this paper,the candidate solution of firefly algorithm is optimized by using local optimal solution,and the variable step size factor is introduced,so as to find the fuzzy factors of interval type-2 fuzzy C-means clustering algorithm adaptively.The interval type-2 fuzzy C-means clustering is carried out combining with the fuzzy factors obtained by optimization and the image change information is extracted by iteratively updating the membership degree.Finally,the weighted Karnik-mendel algorithm based on compound trapezoid rule is used to reduce type and resolve fuzzy to optimize clustering centre.And the change types are determined according to the principle of maximum membership.Through experimental verification,the method in this paper obtains the better fuzzy factors and more accurate clustering centers,has better robustness,and improves the change detection accuracy,and the detected change area is more elaborate.