基于模拟退火算法的植被参数遥感反演

提出了基于模拟退火( SA, Simulated Annealing ) 算法的植被参数( 叶面积指数和叶绿素含量) 反演方案。该方案以冠层反射率模型( SAIL, Scat tering by Arbit rarily Inclined Leav es) 作为正向模型, 分别以Bo ltzman 模拟退火( BSA , Bolt zman Simulated Annealing) 、快速模拟退火( FSA,Fast Simulated Annealing ) 、极快速模拟再退火( VFSA, Very Fast Simulated Anneal ing ) 算法为优化方法, 并采用模型输出的光谱反射率和观测的光谱反射率的残差平方和作为目标函数。模拟反演结果表明: ①模拟退火算法能够跳出局部最优, 得到全局最优解; ②极快速模拟再退火算法在时间效率和反演精度上都优于Bo ltzman 模拟退火和快速模拟退火;③ 在给定的光谱数据没有误差的情况下, 利用模拟退火算法反演SAIL 模型, 能够得到高精度的叶面积指数和叶绿素含量。

A Simulated Annealing Algorithm for Retrieval of Vegetation Parameter from Optical Remote Sensing Data

The optimization approach is one of the most promising methods for retrieval of vegetation parameter from canopy reflectance model based on optical remote sensing data. In this study , a canopy reflectance model ( SAIL, Scattering by Arbitrarily Inclined Leaves) is adopted as forward model and three different simulated annealing algorithms( Boltzman simulated annealing, fast simulated annealing and very fast simulated re-annealing ) are developed as global optimization scheme to simultaneously retrieve leaf area index and content of chlorophy ll, respectively . The Sum of Squared Residuals (SSR) between spectral reflectance by SAIL model and by observation is selected as cost function. The performance of these algorithms is demonstrated with simulated data sets. We can draw following conclusions: ① this algorithm is able to escape local energy minima and can converge to a global energy minimum; ② the very fast simulated re-annealing algorithm priorto Boltzman simulated annealing and fast simulated annealing ; ③under no noise conditions, we can obtain the estimation of leaf area index and chlorophyll content accurately .