BF-FCM Clustering Algorithm And Its Application In The Image Segmentation

In recent years, data mining (DM) has aroused great concern. The aim of DM is to extract the mass of data from all kinds of documents and have it extensively applied, and transfer it into useful information and knowledge. In DM technology, cluster analysis is a widely used method.Fuzzy c-Mean (FCM) clustering algorithm is one of the widely applied algorithms in unsupervised model recognition fields. However, FCM has its own deficiency:before proceeding, the number of clusters must be confirmed, and due to the improper choices of initial cluster-centers, the result is extremely easy to fall into local minimum, etc.This thesis is an intensive study on the bacterial foraging optimization algorithm. It makes an adaptive step change and combines with fuzzy clustering algorithm, using the bacterial foraging optimization algorithm to optimize the criterion functions of FCM, and brings up a combined clustering algorithm (BF-FCM) based on the Fuzzy c-Mean algorithm and the bacterial foraging optimization algorithm. This algorithm combines the global search capability of the bacterial foraging optimization algorithm and the quick local search capability of the FCM algorithm, and utilizes the single bacterium's independent search capability and the mass of bacteria's global search capability, effectively overcomes the FCM's sensitivity to initial value, tendency to fall into local minimum, etc., and meantime, enhances the capability to skip the local minimum. Experimental results show that the new algorithm acquired lower objective values, better clustering evaluation index than traditional FCM algorithm.Another key point of this thesis is the application of the BF-FCM to image segmentation. Image segmentation is the process of extracting objectives or interested area from input images. It is an important step in the process of objective detection and identification. Fuzzy clustering has been widely applied in image segmentation. The new algorithm in this thesis is applied in image segmentation, and is proved to be obviously superior to traditional FCM algorithm in the results of the qualitative and quantitative analysis on the segmentation results of eight groups of images.