Cluster Analysis-Based Image Segmentation And Image Recognise

Image segmentation and image recognise is still a problem now. Because of the intricacy of the world and the multiplicity of approaches to get pictures, there is a tremendous difference among different images and it is difficult to describe images by consistent method and model.Firstly, the paper does research various methods for image segmentation and does classify the methods into three classes that consist of direct segmentation, model-based segmentation and texture segmentation. And then the paper put forward a framework that is used to image segmentation and image recognise and is composed of object space, knowledge expression system and compartmentalized object space. Three parts make up of two mapping. One mapping is from object space to feature space and the other is from feature space to compartmentalized object space. Actually, two mapping are two courses which are feature obtaining and cluster analysis. The course of cluster analysis is a course for compartmentalizing. And equivalence relation is the base of compartmentalizing. The paper does construct equivalence relation by comparability. A example of license segmentation is gived in the paper.Decision tree is a important classifying method and it uses training aggregate to construct instrument like tree to classify modes. Once decision tree is constructed, the classifying becomes a searching process. So the classifying process of decision tree is very fast. Embranchment criterion is very important and pivotal to construct decision tree. After some typical decision tree arithmetic are discussed, the paper puts forward a metric-based decision tree. The decision tree does combine linear classifying instrument with decision tree and does improve effect by decreasing the layer number of tree. At the last, the paper gives experiment data and experiment result for improved decision tree. Experiment result is get by the means of confirmation and intersectional confirmation.