基于蚁群算法的模糊C均值聚类的改进研究

在图像分割的研究中,模糊C均值(FCM)聚类算法较之前的硬聚类有了很大的改进,是一种基于函数最优方法的聚类算法,然而传统的FCM算法的聚类中心及个数难以确定,搜索过程易陷入局部最优。因此,提出一种基于蚁群算法的改进的FCM聚类算法。该算法利用了蚁群算法全局优化特征以及较强鲁棒性的特点,将通过蚁群算法得到的聚类中心及个数应用到传统FCM算法中,弥补了传统FCM聚类算法的不足。该算法对图像进行分块处理,并引入多尺度梯度,提高了图像分割的准确性,最后通过实验验证了该算法的有效性及实用性。

Improved Fuzzy C-means Clustering Based on Ant Colony Algorithm

In the study of image segmentation, fuzzy C-means clustering algorithm (FCM) has been greatly improved compared to the previous hard clustering, which is a clustering algorithm based on a function of best practices. However, the clustering center and number are difficult to be determined for the traditional FCM, also the search process is easy to fall into local optimum. So an improved FCM clustering algorithm is proposed based on the ant colony algorithm. The improved algorithm uses the global optimization features and strong characteristics of robustness of ant colony algorithm. The cluster centers and number obtained by ant colony algorithm are applied to a traditional FCM algorithm to make up for the shortcomings of the traditional FCM. The improved algorithm improves the image segmentation accuracy by processing the image blocks and introducing the multi-scale gradient. Finally the effectiveness and the practicality of the improved algorithm is verified through the experiments.