基于树结构的遗传规划图像分类器,仅能从树根节点输出单一实数值,要完成多类图像分类任务比较困难.提出遗传规划多类图像分类算法,等差权值中心动态边界确定分类算法和权值快速下降中心动态边界确定分类算法.在遗传规划分类器进化过程中,动态确定训练样本各类的边界,通过不同的权重设置策略,使类的边界值设定更趋合理,以提高分类器分类的准确度.图形分类测试表明,这2种算法,用于难度大的多类图像分类问题,能提高分类的准确度,分类效果理想.%Although it is applicable to output a single real value from the root node with the image classifier of genetic programming based on tree structure, it is difficult to cope with more complicated tasks of multi-classification by the same means. Two new approaches of classification algorithm for multi-image classification in genetic programming (GP) are introduced? Including algorithm of centered dynamic range selection with the power value of arithmetic progression (CDRSPVAP) and that of centered dynamic range selection with quick-decreasing power value of arithmetic progression (CDRSQPVAP). During the development of classifiers for GP multi-image classification, different sets of power values are tested to achieve a more suitable range of margin values to improve the accuracy in classification. The tests show that the two approaches can be used for obtaining more accurate results when dealing with complicated problems of multi-image classification, and perform better than using the algorithm of either static or dynamic range selection.
展开▼
