基于冠层尺度的枣树色素含量的高光谱估算模型

发布时间：2019-08-08 07:48

【摘要】：植物冠层色素含量与氮素含量具有高度的相关性,是农业遥感中的关键研究因素。本研究的主要目的是:(1)对比偏最小二乘回归和支持向量机两种建模方法对枣树冠层色素的预测精度;(2)构建基于高光谱数据的枣树冠层色素含量定量反演模型,为枣树冠层色素含量的快速、无损、廉价、环保的测定提供一定的理论依据和技术支持。相关性分析结果表明,枣树冠层色素与高光谱数据之间具有较好的相关性,但叶绿素、叶绿素a要优于叶绿素b和类胡萝卜素。独立样本对模型的预测性能检验结果表明,偏最小二乘回归和支持向量机均能有效的估算枣树色素含量,但不同色素的偏最小二乘回归模型和支持向量机模型的预测精度存在一定的差异,叶绿素和类胡萝卜素的支持向量机模型的预测精度要高于偏最小二乘回归模型,而叶绿素a和叶绿素b则相反。比较不同色素的最佳反演模型的预测精度表明,叶绿素、叶绿素a和类胡萝卜素的预测精度要优于叶绿素b,前三者的决定系数大于0.8,残余预测误差高于2.0,平均相对误差低于13%,而叶绿素b的对应值分别为0.60%,20.79%和1.79%。
[Abstract]:There is a high correlation between plant crown pigment content and nitrogen content, which is a key research factor in agricultural remote sensing. The main purposes of this study are as follows: (1) to compare the prediction accuracy of partial least squares regression and support vector machine for jujube crown pigment; (2) to construct a quantitative inversion model of jujube crown pigment content based on hyperspectral data, which provides some theoretical basis and technical support for the rapid, nondestructive, cheap and environmental protection determination of jujube crown pigment content. The results of correlation analysis showed that there was a good correlation between crown pigment and hyperspectral data, but chlorophyll and chlorophyll a were better than chlorophyll b and carotenoid. The prediction performance test of independent samples to the model shows that partial least square regression and support vector machine can effectively estimate the pigment content of jujube tree, but the prediction accuracy of partial least squares regression model and support vector machine model of different pigments is different. The prediction accuracy of chlorophyll and carotenoid support vector machine model is higher than that of partial least squares regression model, while chlorophyll a and chlorophyll b are the opposite. The prediction accuracy of the best inversion model of different pigments shows that the prediction accuracy of chlorophyll a and carotenoid is better than that of chlorophyll B. the determination coefficient of the first three is more than 0.8, the residual prediction error is more than 2.0, and the average relative error is less than 13%, while the corresponding values of chlorophyll b are 0.60%, 20.79% and 1.79%, respectively.
【作者单位】：塔里木大学植物科学学院;新疆农垦科学院;塔里木大学机械电器化工程学院;
【基金】：国家自然科学基金项目(11564032,41161068) 新疆兵团工业科技计划项目(2014BA019)资助
【分类号】：S665.1;S127

【相似文献】