最小二乘支持向量机用于时间序列叶面积指数预测

Using least squares support vector machines to estimate time series leaf area index

遥感反演的叶面积指数(LAI)时间序列被广泛应用于气候模拟、作物长势监测等研究。但遥感数据受天气等因素影响,时间序列的LAI数据存在缺失。支持向量机(SVM)是一种有效的数据分类和回归预测工具,而最小二乘支持向量机(LS-SVM)是对SVM的有效改进。以西藏那曲县为例,使用2003-2011年MODIS LAI产品,分别用LS-SVM和SVM两种方法对研究区域2011年LAI时间序列进行预测,并用MODIS原始LAI以及部分地面实验样点值进行验证。结果表明,基于LS-SVM的LAI时间序列预测算法的精度比基于SVM的算法高,从而证明LS-SVM方法能够弥补遥感反演时间序列LAI数据的缺失问题,对提高时间序列的LAI遥感产品质量具有重要意义。

The multi-temporal leaf area index (LAI) data retrieved from remote sensing images have been widely used in climate simulation, crop growth monitoring and etc. However,there might be some missing data owing to temporal resolution, weather and some other factors. The support vector machine (SVM) is a kind of machine learning algorithm that has excellent properties. The least squares support vector machine (LS-SVM) algorithm is an improved algorithm of SVM. In this paper, the LS-SVM and SVM models were used to predict the LAI time series products of MODIS data of Naqu in year 2011, based on the MODIS LAI from 2003 to 2011. The results show that LS-SVM method performs better than SVM method. Therefore the predicted LAI data is proved to be very supportive for making up for the loss of remote sensing LAI time-series data, the LS-SVM method proposed in this study is significant to improve the quality of the LAI time series remote sensing products.