摘要
GF-1影像有丰富的光谱和纹理信息,能清楚地反应地物的同时也带来了大量的噪声。由此,文章采用多级多层次分割策略,确定耕地的分割范围,再借助eCognition中最佳分割尺度评价工具ESP获取最优尺度参数,并提取光谱、形状和纹理特征;在此基础上,运用ReliefF算法对分类结果进行优化,并对分类结果进行精度评价。实验结果表明,ReliefF算法能有效地去除无关特征的影响,提高遥感影像分类精度。
GF-1 image has abundant spectral and texture information,which can clearly reflect ground objects and bring a lot of noise.Therefore,in this paper,the using multilevel multi-level segmentation strategy determines the scope of the division of the cultivated land,then with the help of eCognition optimal segmentation scale evaluation tool ESP to obtain the optimal scale parameter,and extracts the spectral,shape and texture characteristics.On this basis,ReliefF algorithm was used to optimize the classification results and evaluate its accuracy.The experimental results show that ReliefF algorithm can effectively remove the influence of irrelevant features and improve the classification accuracy of remote sensing images.
作者
梁加玲
刘彦花
徐军
王茜茜
欧镇丽
LIANG Jia-ling;LIU Yan-hua;XU Jun;WANG Xi-xi;OU Zhen-li(Institute of Surveying and Mapping of Natural Resources,Nanning Normal University,Nanning Guangxi 530001,China;School of Geographic Sciences and Planning,Nanning Normal University,Nanning Guangxi 530001,China)
出处
《地矿测绘》
2020年第3期1-5,共5页
Surveying and Mapping of Geology and Mineral Resources
基金
国家自然科学基金(41461116)
广西自然科学基金“广西旅游流时空流动模式研究”(2020GXNSFAA159065)。
关键词
遥感影像
多尺度分割
RELIEFF算法
影像分类
remote sensing image
multi-scale segmentation
ReliefF algorithm
image classification
