Research On Images Classification Decision Tree Based On Multiple Figures

It's a hot problem in computer vision research area that how toorganize the image data reasonably and effectively and combining theimage low figures how to use artificial intelligent and knowledge discoveryin the images classification.In order to observe and understand word with computer vision welland then accomplish the certain task in some extent intelligence, some basework has done in this paper. It is researched that the theory and method ofimage figure extraction, classification figures choose and rules acquisition.The researches of this paper are as follows:Firstly, color, texture and shape figures of image are discussed. Andthe figures with shift, scale and rotation invariance are used as the initialfigure set.Secondly, two indexes of classification figures choose in decision treealgorithm are analyzed and the figures choose standard based relationshipof figure distribution among various classes is proposed. Through theclassification error definition based on distinct among classes, the conceptsabout classification figure are defined and as the theory basis for thedecision tree generation.Thirdly, it proposes the decision tree generation algorithm based onsamples total distribution, then analyses the performance of algorithm andoptimizes the algorithm. At the same time, it also discusses the influence ofdecision tree with the fluctuation of objects number.Fourthly, it analyses the classification rules of different fields of figurespace. In the process of decision tree generation, the prior knowledge isinducted. And the classification rules extracted method with the weighteddecision tree is proposed.At last, this paper will testify the accuracy of the classifying thoughtand evaluate the error rate, precision and understandability of theclassifying algorithm by a group of images.