多种光谱指标构建决策树的水稻种植面积提取

Extraction of paddy rice area by constructing the decision tree with multiple spectral indices

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NDVI>0.70,0.25<DVI≤0.45,0.53<EVI≤0.80,RVI>5.5且0.30<ρNIR≤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。

Accuracy in discriminating paddy rice can be improved effectively via different spectral indices selected rationally to set the decision tree rules. In the study, multispectral images of 30-m resolution HJ1A and 16-m resolution GF1 taken on August 3 in Huaian city of Jiangsu province were compared. Because of better spectral characteristics to iden-tify different surface objects, HJ1A was selected to determine the threshold ranges of NDVI, EVI, DVI, and RVI, which should satisfy NDVI〉0-7, 0-25〈DVI≤0-45, 0-53〈EVI≤0-8, RVI〉5-5 and 0-3〈ρNIR≤0-46. The threshold of vegetation indices together with reflectance in near-infrared wavelength were used for decision tree classification which was employed for follow-up rice area extration. The extraction accuracy was lower （ 87-29%） by HJ image than by GF image （ 93-7%） , suggesting that it is an effective way to extract paddy rice area using decision tree classification model based on multispectral images.