高光谱数据预处理对大豆叶绿素密度反演的作用

发布时间：2018-06-25 08:40

本文选题：高光谱 + 预处理　；参考：《遥感信息》2017年04期

【摘要】：针对高光谱数据异常值影响叶绿素密度反演精度的问题,以大豆叶片为研究材料,利用马氏距离和蒙特卡洛交叉验证法(Monte Carlo cross validation,MCCV)剔除异常样本,探讨13种高光谱数据预处理方法对叶绿素密度偏最小二乘法(partial least square,PLS)建模的影响。结果表明,马氏距离法和MCCV剔除异常样本能提高校正模型的精度,在权重系数为1时剔除异常样本数3个,模型精度最高,校正集决定系数和均方根误差分别为0.821和0.112。微分处理能大幅度提高模型的预测精度,二阶微分处理效果最好,校正集决定系数和均方根误差分别为0.998和0.012,验证集决定系数和均方根误差分别为0.961和0.139,具有比原始光谱更高的精度。因此,适宜的高光谱数据预处理可有效提高大豆叶绿素密度估测精度。
[Abstract]:Aiming at the problem that the abnormal value of hyperspectral data affects the accuracy of chlorophyll density inversion, the abnormal samples are eliminated by using Markov distance and Monte Carlo cross validation (MCCV) method. The effects of 13 preprocessing methods of hyperspectral data on chlorophyll density partial least square (partial least) modeling were investigated. The results show that the method of Markov distance and MCCV can improve the accuracy of the correction model. When the weight coefficient is 1, the number of abnormal samples is 3, the model accuracy is the highest, the determination coefficient of correction set and the root mean square error are 0.821 and 0.112, respectively. Differential processing can greatly improve the prediction accuracy of the model, and the second-order differential processing is the best. The correction set determination coefficient and root mean square error are 0.998 and 0.012, respectively, and the validation set decision coefficient and root mean square error are 0.961 and 0.139, respectively, which have higher accuracy than the original spectrum. Therefore, suitable hyperspectral data pretreatment can effectively improve the accuracy of soybean chlorophyll density estimation.
【作者单位】：四川农业大学农学院;农业部西南作物生理生态与耕作重点实验室;四川省作物带状复合种植工程技术研究中心;
【基金】：国家重点研发计划项目(2016YFD030060203) 国家自然科学基金(31571615)
【分类号】：S565.1;TP79

【相似文献】