Large-scale Dynamic Monitoring Of Rice Growth Based On Parallel Algorithm Assimilation Of Remote Sensing And Crop Growth Model

Assimilation of remote sensing crop growth model is an effective way to obtain continuous dynamic information of regionalized vegetation growth,which has been widely used in agriculture.However,the efficiency of assimilation technology can not meet the need of continuous dynamic monitoring of large-scale crop growth when large-scale extension is carried out.Therefore,to improve the assimilation efficiency of remote sensing crop growth model is of great significance for fast and accurate acquisition of regional scale vegetation growth dynamic information,and can effectively guide agricultural production.In this study,Zhuzhou City,Hunan Province,which is an important agricultural and industrial base in China,was selected as the research area.Landsat and HJ-1 remote sensing image data,crop parameters and meteorological data were collected.A remote sensing/crop growth model assimilation framework was constructed by selecting the appropriate input parameters.A parallel assimilation program was successfully developed on the platform of CDUA of GPU,and the assimilation efficiency of serial program and parallel program was compared.Spatio-temporal simulation of the growth parameters in the growing season was carried out,and the growth trend of rice from 2009 to 2016 was analyzed.The main research work and conclusions of this paper are as follows:(1)A remote sensing/WOFOST model Assimilation Algorithm Based on PSO was constructed,in which environmental stress factors were taken as input parameters to be optimized.Taking environmental stress factors as input parameters to be optimized,the degree of rice affected by environmental stress in each pixel in the study area can be accurately obtained,and more accurate rice growth information can be obtained under the influence of environmental stress.(2)A remote sensing/WOFOST model assimilation program based on PSO algorithm is developed on the platform of GPU CUDA.The results show that the assimilation efficiency of parallel assimilation program is about 280 times higher than that of serial program for 401296 rice pixels in the study area using P40 GPU CUDA platform.(3)Based on reliable and efficient parallel assimilation results,the temporal and spatial analysis of rice growth in 2009-2016 was carried out.The results showed that the growth of rice in 2014 was poor,and environmental stress had a greater impact on the growth process of rice;the growth of rice in 2009 and 2013 was better,but the impact of environmental stress was less;the growth of rice in other years was in the middle level.On the premise of ensuring the accuracy of assimilation,a parallel Assimilation Algorithm Based on GPU CUDA platform has been adopted to greatly improve the assimilation efficiency of remote sensing crop growth model,thus realizing fast and accurate acquisition of long time series and large regional scale rice growth information.This method has important application value in the field of agricultural remote sensing.