以提高遥感影像分类精度为研究目的,以Matlab 7.0作为实验平台,利用7种单分类器分别在测量级和抽象级上进行多分类器结合的方法研究,并进行实验验证.其中针对在非监督分类中后验概率计算时的类别中心确定问题,本文在分类结束后重新计算各个类别特征的平均值作为类别中心.本文还对投票法进行了改进,通过比较各个类别后验概率之和的方法解决多个类别同时为得票最高类别时的像元类别归属问题,并对实验结果进行精度评价和分析.结果表明多分类器结合方法对单分类器的总体分类精度有一定程度的提高.%In order to improve the classification accuracy,this paper performed multiple classifier integration at abstract level and measurement level by using MATLAB 7. 0 software,and seven classifiers were studied. To determine the class center when calculating posteriori probability in unsupervised classification,the average of characteristics was recalculated as the class center after classification. The method of voting has been improved. By comparing the sum of the posterior probability of each class, the problem of determining to which class a certain pixel belongs when more than one class win the highest number of votes was also solved. By accuracy assessment,it showed that the proposed method can effectively improve the overall classification accuracy.
展开▼
