摘要
在地物环境状况较为复杂时传统基于统计学遥感分类算法难以得到较高的分类精度。针对这一问题,这里将基于分组策略的改进基因表达式编程算法(BS-GEP)应用到遥感图像分类问题中,避免传统的基因表达式编程算法由于种群多样性破坏引起局部收敛,解决地物状况复杂时难以得到较高分类精度的问题。实验结果表明:基于分组策略的基因表达式编程算法的分类器提取的分类规则能转为数学表达式形式并能获得较高的分类精度,与基因表达式编程算法(GEP)相比分类结果混淆程度相对较低,与最大似然法相比分类结果相对清楚,模型分类精度达到93%。
It is difficult for the traditional statistical remote sensing classification algorithm to get higher classifica- tion accuracy under the condition of complex ground state. To solve this problem, BS-GEP algorithm was intro- duced to the study of remote sensing image classification problems in this paper, to avoid local convergence of the algorithm caused by the population diversity, the characteristic of the traditional GEP, and solve the problem of getting higher classification accuracy difficultly under the complex condition ground state. The experimental results have shown that classification rules based on the BS-GEP classifier can be converted into mathematical expressions and obtain higher classification accuracy. Compared with GEP algorithm, the confused degree of the classification results are relatively low, and compared with maximum likelihood algorithm, the classification results are relatively clear. The classification accuracy of the classifier has been reached to 93%.
出处
《测绘科学技术学报》
CSCD
北大核心
2015年第4期401-404,共4页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目(41373101)
关键词
遥感图像分类
基因表达式编程
局部收敛
分类规则
分类精度
remote sensing image classification
gene expression programming
local convergence
classification rules
classification accuracy
