基于遗传算法优化决策树的多光谱影像分类研究

The classification of multi-spectral images based on genetic algorithm and diction tree

决策树C4.5算法采用局部贪婪搜索的策略,会增加局部最优解的风险;在样本有限的情况下,所产生的分类规则会过于依靠样本,往往造成决策树不能有效地挖掘出有价值的分类规则和形式。本文将决策树算法用到基于像斑的多光谱分类研究中,尝试引入遗传算法,对决策树分类规则进行优化。试验结果表明,在样本有限的情况下,该方法比单个决策树具有更高的分类精度。

Decision Tree C4. 5 algorithm using greedy local search strategy may increase the risk of local optimal solution. In the case of limited samples, the classification rules depend much on samples, which lead to a result that the decision tree can not mining valuable classification rules correctly. The paper uses the decision tree to the classification of multi-spectral images based on the homogeneous region, and tries to introduce genetic algorithm to optimize the classification rules derived by the decision tree. The result shows that this method can distinguish objects accurately and improve the precision.