多种光谱指标构建决策树的水稻种植面积提取
发布时间：2018-01-17 23:35
本文关键词：多种光谱指标构建决策树的水稻种植面积提取 出处：《江苏农业学报》2016年05期 论文类型：期刊论文
【摘要】：合理选取不同光谱指标制定决策树规则,能有效提高决策树分类法提取水稻面积的精度。本研究以江苏省淮安市为例,选取30 m空间分辨率HJ1A和16 m空间分辨率GF1多光谱影像,在对不同地物样点像元光谱特征分析的基础上,选择地物光谱特征明显的GF影像计算NDVI、EVI、DVI和RVI,并提取影像近红外波段反射率,利用上述5种光谱指标确定不同地物分类阈值来对两景影像进行决策树分类,进而获取淮安市水稻面积和分布情况。结果表明,GF影像地物光谱特征较明显,有利于识别不同地物,可用来确定基于多种光谱指标分类的阈值范围。其中,水稻判别条件为NDVI0.70,0.25DVI≤0.45,0.53EVI≤0.80,RVI5.5且0.30ρNIR≤0.46。HJ影像和GF影像提取水稻面积的样本精度分别为87.29%和93.70%,GF影像比HJ影像的水稻面积提取精度提高了6.41个百分点,说明利用多种光谱指标构建决策树分类模型是一种有效提取水稻种植面积的方法。
[Abstract]:Reasonable selection of different spectral indicators to make decision tree rules can effectively improve the precision of rice area extraction by decision tree classification. This study takes Huaian City Jiangsu Province as an example. The 30 m spatial resolution HJ1A and 16 m spatial resolution GF1 multispectral images are selected, and the spectral characteristics of the pixels of different ground objects are analyzed. Select GF image with obvious spectral characteristics to calculate DVI and RVI, and extract the near infrared reflectance of the image. The above five spectral indexes were used to determine the threshold of different ground objects classification to classify the two scene images, and then to obtain the rice area and distribution in Huai'an City. The results showed that the spectral characteristics of ground objects in GF images were obvious. It can be used to determine the threshold range of classification based on various spectral indexes, and the rice discriminant condition is NDVI 0.70 / 0.25DVI 鈮，
本文编号：1438507
本文链接：https://www.wllwen.com/kejilunwen/nykj/1438507.html
