基于粒子滤波的LAI时间序列重构算法设计与实现

Li Manman, China Agricultural University, Beijing , College of Information and Electrical Engineering; Liu Junming, China Agricultural University, Beijing , College of Information and Electrical Engineering; Wang Pengxin, China Agricultural University, Beijing , College of Information and Electrical Engineering

Design and realization of reconstructing LAI time-series data by particle filter | 基于粒子滤波的LAI时间序列重构算法设计与实现

2012

Li Manman, China Agricultural University, Beijing (China), College of Information and Electrical Engineering | Liu Junming, China Agricultural University, Beijing (China), College of Information and Electrical Engineering | Wang Pengxin, China Agricultural University, Beijing (China), College of Information and Electrical Engineering

中国人. 遥感观测的叶面积指数（LAI）时间序列数据广泛应用于作物长势监测，但数据受大气条件等影响，存在数值偏低和时间序列数据缺失等问题。为此，本文设计了一种基于重采样粒子滤波的LAI时间序列重构算法，以LAI为同化变量，在WOFOST模型本地化的基础上，实现了遥感LAI数据和WOFOST模型模拟的LAI数据的同化，以重构LAI时间序列。算法将WOFOST作物模型简化为LAI状态随时间演变的非线性计算方程，作为重采样粒子滤波的状态转移方程；将地面实测LAI数据和遥感LAI数据建立的线性方程，作为重采样粒子滤波的观测方程，建立LAI时间序列数据同化模型。以带权重粒子表示LAI时间序列状态后验分布，并在循环迭代中对粒子重采样，以此实现单点和区域LAI时间序列重构。应用该算法，对河北省冬麦区2010年LAI时间序列进行重构，结果表明，基于重采样粒子滤波的LAI时间序列重构算法在单点和区域上得到的LAI值明显更接近冬小麦实际生长状况，且算法能够弥补遥感LAI时序数据的缺失，为进一步的作物长势监测提供基础支撑。

显示更多 [+]

英语. Leaf area index (LAI) time-series data retrieved from remote sensing images have been widely used in monitoring crop growth. However, affected by atmosphere and other conditions, the data might be underestimated and even missing. Therefore, an algorithm was designed in this paper to reconstruct the remote sensing LAI time-series data by particle filter. Particle filter was used for the reconstruction. LAI was used as a variable to localize WOFOST model. Reconstruction of LAI time-series data can be done through assimilation of the remote sensing LAI data and WOFOST-LAI data. The algorithm simplified WOFOST model to be the nonlinear evolution of LAI changing with time, which was used as the state transition equation of re-sampling particle filter. And it got the observation equation with measured LAI data and the remote sensing LAI time-series data, to establish assimilation model of LAI time-series data. And the weighted particles represented the posterior distribution of LAI, then re-sampled the particles in the iteration in order to reconstruct LAI time-series data at the point and in the region. We reconstructed the LAI time-series data of Hebei Province in 2010 with this algorithm. According to the results, we could get better LAI time-series data at the point and in the region with this particle filter-based algorithm, which was closer to the actual growth conditions of crop. And it could also make up for the loss of remote sensing LAI time-series data. Therefore, the reconstructed LAI data could be very supportive for monitoring crop growth.

显示更多 [+]