摘要
为强化高光谱成像技术在近地农业方面的应用,以农田近红外高光谱图像为研究对象,利用高光谱成像技术,结合光谱分析方法和监督分类方法,对农田图像进行分类。针对高光谱图像数据量大、非线性等特点,采用主成分分析(PCA)和支持向量机(SVM)法建立农田图像分类器。在利用光谱信息分类的基础上,采用空谱一体化方法对光谱分类结果进行修正,去除孤立点和噪声的影响。基于支持向量机的总体分类精度为88.4%,采用空谱一体化方法的总体分类精度最高达89.7%,说明利用空间信息修正光谱信息可以提高近地农田对象的分类精度,为基于高光谱图像的近地农田识别提供理论依据。
In order to intensify the application of hyperspectral imaging technology in near field agriculture,near-infrared hyperspectral images were selected as research objects,and the hyperspectral imaging technology combining with spectral analysis method and supervised classification method was used to classify the farmland images. Since the hyperspectral data had the characteristic of huge and nonlinear,principal component analysis（ PCA） and support vector machine（ SVM） were adopted for classifier establishing. On the basis of spectral classification,spatial-spectral integration method was used to amend the spectral classification results,the isolated points and noise were removed. The results showed that the overall classification accuracy by SVM could reach 88. 4%,and the highest overall classification accuracy by spatialspectral integration method was up to 89.7%,indicating that using spatial information to modify spectral information could improve the classification accuracy of farmland objects,which would provide theoretical basis for hyperspectral image identification of near field farmland.
作者
苗荣慧
黄锋华
杨华
邓雪峰
陈晓倩
MIAO Rong-hui;HUANG Feng-hua;YANG hua;DENG Xue-feng;CHEN Xiao-qian(College of Information Science and Engineering,Shanxi Agricultural University,Taigu 030801,China;College of Information Engineering,North west A＆F University,Yangling 712100,China)
出处
《江苏农业学报》
CSCD
北大核心
2018年第4期818-824,共7页
Jiangsu Journal of Agricultural Sciences
基金
国家自然科学基金项目(31671571)
山西农业大学青年科技创新基金项目(2017013)
关键词
高光谱图像
空谱一体化
农田图像分类
主成分分析
支持向量机
hyperspectral image
spatial-spectral integration
farmland image classification
principal componentanalysis （PCA）
suppoa vector machine （SVM）
