中文摘要：

以云南省香格里拉县建塘镇的高山松为研究对象,使用ASD Field Spec 3便携式野外地物光谱仪测定高山松叶片光谱,并在实验室测定叶片样本的叶绿素含量。经光谱分析技术及统计相关分析法进行光谱数据的分析处理,提取原始光谱及一阶微分光谱特征波段和光谱特征变量,建立与叶绿素含量间的单变量估测模型和双隐层BP神经网络预测模型,并采用决定系数（R2）、均方根差（RMSE）和相对误差（RE）进行精度检验。结果表明,单变量模型以一阶微分光谱反射率的三次函数模型为最优模型,其R2、RMSE、RE分别为0.511、1.297 6mg/g、10.06%,而基于双隐层BP神经网络最优模型的R2、RMSE、RE分别为0.637、0.384 1mg/g、9.47%,精度达到90.53%,经比较得出其具有较优的预测能力,充分体现BP模型的可行性,为快速、准确的估测高山松叶绿素含量提供有利的理论依据。

英文摘要：

Pinus densata occurring in Jiantang township of Shangrila,Yunnan Province was taken as the re- search object,the spectral reflectance of the leaves were measured by a portable field surface object spec- trometer （model ASD Field Spec 3） ,while the content of chlorophyll in the leaves were measure in labora- tory. The spectral data were processed using spectral analysis techniques and statistical methods to extract the original spectrum, differential spectral and spectral characteristic variables, which were then adopted to establish the single variable estimation models and the double hidden layer BP neural network prediction model estimate chlorophyll content. The fitness between the predicted values and the measured values were evaluated by the determination coefficient （R2） , the root mean-square error （RMSE） and the average rela- tive error （RE）. The results showed that the best single variable model was the cubic model using the first order differential spectral reflectance as a variable, in which the values of R2 ,RMSE and RE were 0.511,1. 297 6 mg/g and 10.06% ,respectively. The values of R2 ,RMSE and RE of the best double hidden layer BP model were 0. 637,0. 3841 mg/g and 9.47 %, respectively, in which the prediction accuracy was 90.53 %, in- dicating it had better predictive ability,and could provide a rapid and accurate way to estimate the content of P. densata chlorophyll leaves.