基于PLS算法与遥感影像区域小麦实际单产估测.docVIP

基于PLS算法与遥感影像区域小麦实际单产估测 　　摘要 为进一步提高遥感估产精度，显示国产影像在农业估产中的应用效果，该研究以2010―2013年HJ-1A/1B影像为遥感数据，分析了卫星遥感变量与小麦实际单产的定量关系，运用偏最小二乘回归算法构建及验证了以实际单产为目标的多变量遥感估产模型。研究表明：实际单产与所选用的大多数遥感变量间关系密切，且多数遥感变量两两间具有严重的多重相关关系；实际单产偏最小二乘回归模型的最佳主成分为5，且植被衰减指数、绿色归一化植被指数、调整土壤亮度的植被指数、比值植被指数和归一化植被指数为实际单产遥感估测的敏感变量；建模集和验证集实际单产估测模型的决定系数分别为0.74和0.70，均方根误差分别为754.05、748.20 kg/hm2，相对误差分别为11.5%和 8.88%，且估测精度比线性回归算法分别提高20%以上和40%以上，比主成分分析算法分别提高18%以上和30%以上，说明偏最小二乘回归算法模型估测区域实际单产的效果要明显好于线性回归和主成分分析算法，该模型应用结果与小麦实际单产区域分布情况相符合，为提高区域小麦实际单产的遥感估测精度提供了一种途径。 　　关键词 遥感；产量估测；偏最小二乘法；小麦 　　中图分类号 TP79 文献标识码 A 文章编号 1007-5739（2015）18-0209-02 　　Estimation of the Actual Yield Per Unit Area of Wheat Based on PLS Algorithm and Remote Sensing Image 　　LIU Xiao-feng 　　（Seed Management Station of Bayinguoleng Mongolia Autonomous Prefecture of Xinjiang Uygur Autonomous Region，Korla Xinjiang 841000） 　　Abstract In order to further improve the accuracy of crop-yield assess by remote sensing，according to domestic imaging applications in agriculture yield estimation results，taking HJ-1A/1B from 2010 to 2013 as remote image sensing data，satellite remote sensing variable and real output of wheat was analyzed.The study used partial least squaress regression algorithm was built and verified by the actual yield as the target of multivariable remote sensing yield estimation model.The research showed that the actual yield had close relationship with remote sensing variables，and most remote sensing variables between the two had severe multiple correlation relationship；Actual yield of partial least-squares regression model best principal component was 5，and the attenuation of vegetation index，green normalized difference vegetation index，adjust the brightness of vegetation index，soil ratio vegetation index and normalized difference vegetation index were sensitive variables to the actual yield remote sensing estimation；Modeling set and validation set models for predicting the actual yield decision coefficient were 0.74 and 0.70，root mean square error