Based On Multi-agent And Mathematical Morphology Combined Image Clustering

Clustering analysis, also known as segmentation, is an unsupervised learningprocess, Clustering analysis divides the data set according to the requirementthat the similarity between inner cluster classes is higher while between outercluster classes is lower. Clustering analysis is an important research area in datamining, which determines the final mining results and quality. The spatial dataare complex and changeful, and the quantity is tremendous, so the data analysiswork is onerous. Utilizing distributed computing and artificial intelligencetechnology to improve the analysis capabilities of the traditional data miningalgorithms is a research hotspot in the current data analysis field. Automaticclustering of the integration of multi-agent and mathematical morphologyapproach to dealing with complex images is a challenging problem and has ahigh practical value.Focusing on the premise of the local model, Multi-agent technology provides anew solution approach for complex problems, using the advantage of distributedparallel computing, the basic idea of Multi-agent technology is that simple agentthrough the cooperation presents the intelligent behavior characteristics.The clustering analysis algorithm studied in this paper is applied in digitalelevation images (DEM) and medical images. The main work and innovation aresummarized as follows:(1) Based on relevant literature at home and abroad about Multi-Agenttechnology clustering method are deeply analyzed. This paper reviews the basicmethods of clustering analysis, and at the account of the characteristics ofgray-scale image, expounds the feasibility and necessity of intelligent clusteringbased on multi-agent and mathematical morphology.(2) A clustering algorithm named AMMC was proposed, which is based onmulti-agents and gray-scale mathematical morphology operations is proposed..Agent technology and gray-scale mathematical morphology model are combinedin the algorithm. The structural element of the mathematical morphology isselected as agent. Based on the values of Moore Neighborhood or VN Neighborhood in the environment of their spatial location, the agents chooseOCC operator autonomously to do gray dilation or erosion operation toimplement spatial clustering. This algorithm has the capabilities of distributedparallel computing and initial self-analysis. The experimental results show thatAMMC algorithm used in the spatial data mining to discover clusters witharbitrary shape has significant accuracy, reliability, flexibility.(3) Considering the spatial heterogeneity of image, the AMMC algorithm isimproved by using different types of agents. agents are randomly distributed ona discrete spatial grid, and controlled by a same drive-clock agents to choose theappropriate operation with the given ONA operator according to the type of itsstructural, and then agents move to a random clustering spatial location of itsMoore Neighborhood, in the migration process of all agents makes clusteringautomatically generated. The algorithm does not require any priori knowledgeand pre-processing operations, and the initial clustering point is not sensitive totime and without prior input the number of clusters. This algorithm hasdistributed parallel computing capability and preliminary independent analysiscapability. General clustering method directly applied to medical images are notnecessarily the ideal segmentation results of the experimental data for brain MRimages, experimental results show that this algorithm can cluster accurately indifferent organizational structures in medical images with more reliability andflexibility.