基于自适应蚁群的FCM聚类优化算法研究

Optimization of Fuzzy C-Means Clustering Based on Adaptive Ant Colony Algorithm

模糊C均值聚类算法在算法初始化时需要人为设定聚类类别数、随机初始化聚类中心,致使该算法容易陷入局部最优值。为解决此类问题,在蚁群算法中引入信息素更新机制,使其输出的聚类中心更具全局优化的特征和较强鲁棒性的特点;用蚁群算法得到的聚类中心来初始化FCM算法的聚类中心,解决了FCM算法对初始聚类中心敏感的问题;使用结合熵信息与数据几何结构的聚类有效性评价方法对FCM算法和优化FCM算法进行评价,评价结果表明优化的FCM算法性能更优。在仿真实验中,利用提出的优化算法和FCM算法对自然图像、纹理图像和SAR图像进行分割实验,从图像分割的准确性和算法的实时性做对比实验,验证了优化算法的有效性。

The fuzzy C-means（FCM） clustering algorithm needs to artificially set the number of clusters when initializing the algorithm, and randomly initializes the clustering center, which makes the FCM algorithm easily fall into the local optimal solution. The optimized FCM clustering algorithm was used to solve the above problems. The adaptive ant colony algorithm created the initial clustering center and the number of the centers, the data were used as the input of FCM clustering algorithm. The FCM algorithm and the optimized FCM algorithm were evaluated by clustering validity evaluation method of entropy and data geometric structure. Experimental results and analysis show the effectiveness of the method.