摘要
本文基于Landsat 7 ETM+全色图像,以浙江省临安市大峡谷镇为研究区,进行基于遗传算法的遥感图像纹理特征选择的研究.首先利用灰度共生矩阵法和灰度-梯度共生矩阵法对研究区遥感图像进行纹理特征提取,共得到23个纹理特征,然后利用遗传算法对这23个纹理特征进行纹理特征选择,最后得到一组最优纹理特征集.实验结果表明,遗传算法因其自适应性、并行性、能较好地处理大规模复杂数据,且特别适合于解决多目标优化问题等诸多特性,所以是解决特征选择问题的理想方案.
In this paper, an approach to texture feature selection in ETM4-panchromatic image based on genetic algorithms is described. Daxiagu town of Linan county, Zhejiang province is taken as the research area. Firstly, the texture features were extracted from the ETM^pan image by using gray co-occurrence matrix and gray-gradient co occurrence matrix, and all the 23 texture features were obtained. Then texture feature selection was carried out by genetic algorithms. The optimal texture sub-set could be got finally. The results show that the method of genetic algorithms can deal with the large and complicated data because of its characteristics of self-suitability and parallel, and it is very suitable for solving the problem of multi-object optimization, so the method of genetic algorithms is a better way to solve the problem of feature selection.
出处
《南京大学学报（自然科学版）》
CAS
CSCD
北大核心
2008年第3期310-319,共10页
Journal of Nanjing University（Natural Science）
关键词
纹理特征提取
遗传算法
纹理特征选择
LANDSAT7
ETM+
全色图像
texture feature extraction, genetic algorithms, texture feature selection, Landsat 7 ETM 4, panchromatic image, Lin ＇ an city
