红外与毫米波学报, 2009, 28 (5): 342, 网络出版: 2010-12-13
基于可见光/近红外光谱技术的倒伏水稻识别研究
DISCRIMINATION OF LODGED RICE BASED ON VISIBLE/NEAR INFRARED SPECTROSCOPY
稻飞虱 穗颈瘟 可见光/近红外光谱反射率 主成分分析 支持向量分类机 Key words: rice planthopper rice panicle blast visible/near infrared (VIS/NIR) spectral reflectan principal compo- nent analysis (PCA) support vector classification (SVC)
摘要
运用可见光/近红外光谱仪获取正常的和受稻飞虱、穗颈瘟危害而倒伏的水稻冠层光谱反射率, 采用主成分 分析(PCA)方法对反射率光谱进行降维处理, 提取2个主分量光谱.其中, 第一主分量PC1代表了水稻冠层的光谱 特性, 第二主分量PC2反映了倒伏水稻的冠层光谱变化信息.将前2个主分量作为支持向量分类机(SVC)的输入 向量, 建立分类模型.结果表明, 对受稻飞虱危害倒伏的水稻验证数据的识别精度为100%, 对受穗颈瘟危害倒伏的 水稻验证数据的识别精度为90.9%.研究表明可见光/近红外光谱可能是一种有效的倒伏水稻识别方法.
Abstract
Hyperspectral reflectances of the healthy and lodged rice caused by rice planthopper and rice panicle blast were measured with visible/near-infrared (VIS/NIR) spectroradiometer at the canopy level. The principal component analysis (PCA) was used to obtain the principal components (PCs) and to reduce the spectral dimensions of hyperspectral reflec- tance. Two principal components were extracted. The first (PC1) and second (PC2) reveal the general feature of rice spectral reflectance and spectra change of lodged rice relative to healthy rice, respectively. The front two PCs entered the support vector classification (SVC) as the input vectors to build the discrimination model. The recognition accuracies of healthy and lodged rice are 100% and 90.9% for the rice planthopper and rice panicle blast stresses, respectively. The results demonstrate that visible/near-infrared spectroscopy technique may provide potential discrimination accuracy for lodged rice.
刘占宇, 王大成, 李波, 黄敬峰. 基于可见光/近红外光谱技术的倒伏水稻识别研究[J]. 红外与毫米波学报, 2009, 28(5): 342. LIU Zhan-Yu, WANG Da-Cheng, LI Bo, HUANG Jing-Feng. DISCRIMINATION OF LODGED RICE BASED ON VISIBLE/NEAR INFRARED SPECTROSCOPY[J]. Journal of Infrared and Millimeter Waves, 2009, 28(5): 342.