基于Sentinel-1A双极化时序数据的甘蔗株高反演方法

甘蔗株高为甘蔗品种与土壤、气象、水文等因素的综合反映,是甘蔗长势监测与估产的重要指标。研究以华南地区气候与天气条件为基础,通过对覆盖甘蔗全生长期的23景时间序列Sentinel-1A数据进行预处理、矩阵转换与Cloude-Pottier分解,求得双极化雷达植被指数（Dual-pol radar vegetation index, DPRVI）。分析了该指数与甘蔗长势参数（株高）随甘蔗不同生长期的动态变化规律。采用4种经典的经验回归模型（线性、二次多项式、指数、对数）,以分段函数形式对不同生长期的甘蔗株高进行反演,建立最佳反演模型。实验结果表明,拟合模型在分蘖期前相关性最高,二次多项式模型拟合效果最优,决定系数R²与均方根误差分别达到了0.882与0.118 cm,对反演效果最好的分蘖期之前的二次函数模型进行验证,结果表明决定系数R²达0.839,平均绝对偏差为7.4%,说明DPRVI反演甘蔗株高是有效的。将DPRVI与其他3种经典的反演参数进行对比,结果表明,DPRVI的性能优于其他3种参数。通过分析可得,DPRVI可以较好地反演甘蔗生长前期...

Inversion Method of Sugarcane Plant Height Based on Sentinel-1A Dual-polarization Time Series Data

Sugarcane plant height is a comprehensive reflection of soil, meteorology, hydrology and other factors, and it is also an important index for sugarcane growth monitoring and yield estimation. Based on the climate and weather conditions in South China, the dual-pol radar vegetation index (DPRVI) was obtained by preprocessing, matrix transformation and Cloude-Pottier decomposition of the time series Sentinel-1A data of 23 landscapes covering the whole growth period of sugarcane. The dynamic changes of this index and sugarcane growth parameters (plant height) with different growth periods of sugarcane were analyzed. Then four classic empirical regression models (linear, quadratic polynomial, exponential and logarithmic) were used to invert the height of sugarcane plants in different growth periods in the form of piecewise function to establish the best inversion model. Through the experimental results, it can be found that the fitting model had the highest correlation before the tillering stage, and the quadratic polynomial model had the best fitting effect. The determination coefficient R2 and the root mean square error were 0.882 and 0.118cm, respectively. Then the quadratic function model with the best inversion effect before tillering stage was verified and analyzed. The results showed that the determination coefficient R2 was up to 0.839, and the average absolute deviation was 7.4%, which indicated that the model could better guarantee the accuracy of predicting plant height by using DPRVI. Finally, DPRVI was compared and analyzed with other three classical remote sensing parameters. The results showed that the performance of DPRVI was better than that of the other three parameters. Through the analysis, it was concluded that DPRVI can better invert the change of plant height in the early growth stage of sugarcane, and the parameters of plant height can be used as a reference for agricultural management departments.