摘要
遥感图像分类是遥感数据转为信息的重要途径,本文将地统计学和遗传算法结合,构建了一种基于Davies-Bouldin系数和地统计学变异函数的遗传优化指标,实现了基于地统计学纹理约束的启发式遥感图像非监督分类。通过对实验区TM数据的应用,总体分类精度可到达92.54%,与传统遗传算法相比总体分类精度提高5.64%。
Remote sensing image classification is an important way to convert the remote sensing data into useful information.This paper combined geostatistics and genetic algorithm,constructing agenetic optimization index combined Davies-Bouldin Index with geostatistics variogram,and came up with a heuristic and unsupervised classification of remote sensing image constrained with geostatistics texture.Through the application of TM data in experimentation area,it was found that the overall classification accuracy can reach 92.54%,and the overall classification accuracy improves of 5.64% compared with the traditional genetic algorithm.
出处
《遥感信息》
CSCD
2013年第6期13-18,共6页
Remote Sensing Information
基金
2012年南京信息工程大学大学生创新训练项目(12CX034)
国家自然科学基金项目(41001288
41201461)
江苏省自然科学基金项目(BK2010571)
江苏省高校自然科学研究项目(10KJB170010)
关键词
图像分类
遗传算法
地统计学
遥感图像
image classification
genetic algorithm
geostatistics
remote sensing image
