摘要
为实现玉米叶片叶绿素含量的快速无损测定,采集不同氮素水平的玉米叶片,制备校正集样本60个,验证集样本16个,获取400～1 100 nm波段范围的高光谱数据和相应叶绿素含量.采用变量标准化、13点平滑、一阶导数3种预处理方法结合,根据相关系数图谱选择470～760 nm波段作为光谱数据分析对象;利用最小二乘支持向量机建立玉米叶片叶绿素含量与高光谱数据的定量分析模型,基于交叉验证的网格搜索寻找LS-SVM的最优参数,建立LS-SVM模型;所建立的校正模型相关系数为0.96,验证相关系数为0.93.研究结果为高光谱技术在精准减量施肥遥感检测中的应用提供了技术基础.
For the rapid and non-destructive detection of chlorophyll content in corn leaves, representa- tive corn leaves with different N levels were collected. 60 calibration samples and 16 validation samples were prepared. Hyperspectral images in the range of 400 - 1 100 nm were collected and relevant chloro- phyll content was measured according to the National Standard. Standard normalized variation, 13 points smoothing, and first derivative were applied as pretreatment method. According to the correlation coeffi- cient, the wave band of 470 -760 nm was selected as analysis object. Least squares-support vector ma- chines were used to establish the model between the corn leaves＇ chlorophyll content and the hyperspec- tral data. The optimal parameters of LS - SVM were obtained by application of grid-search based on cross- validation. The results of LS - SVM model indicate technical support for hyperspeetral application in re- mote sensing, with correlation coefficient of 0. 96 and calibration coefficient of 0.93, respectively.
出处
《江苏大学学报（自然科学版）》
EI
CAS
北大核心
2011年第2期125-128,174,共5页
Journal of Jiangsu University：Natural Science Edition
基金
国家"十一五"科技支撑计划项目(2007BAD89B04)
关键词
玉米叶片
叶绿素含量
高光谱成像
最小二乘支持向量机
检测
corn leaves
chlorophyll content
hyperspectral imaging
least squares-support vector machines
detection
