Study On Extraction Method Of Rice Planting Area In Cloudy Rain And Fog Area Based On Multi-source Satellite Remote Sensing Data

Paddy rice is one of the main food crops in China,and its output has always ranked first,which has a decisive impact on Chinese food security.Therefore,the extraction of large-scale rice planting area and the acquisition of yield information are of great significance to the scientific guidance of rice production and the rational allocation of water resources.Due to the wide area involved in rice cultivation,the differences between the regions are also large,and it has a strong seasonality.Only through traditional methods such as ground surveys and statistics to obtain rice cultivation information,there are low efficiency and high cost.problem.The emergence of satellite remote sensing technology provides an effective solution to this problem.Compared with traditional statistical methods,remote sensing technology has the advantages of low cost,large range,and high spatial and temporal resolution.However,southwestern China is in a cloudy and foggy area,and single optical satellite remote sensing data is susceptible to cloud and fog.It is the main method to realize the extraction of rice planting information in cloudy and foggy areas.This thesis takes eight provinces(cities,autonomous prefectures)and other provinces in southwestern China as research areas,using MODIS,Landsat,and Sentinel satellite data,and using different research methods to extract rice planting area.The main work and research results completed in this paper are as follows:(1)Based on the MODIS data of time series,the spatial distribution of rice planting and the extraction of rice planting area in South China and southwest China were realized.In this paper,the spatiotemporal filtering method is used for MODIS(moderate resolution imaging Spectral radiometer)is used to fill in the data,and then based on the rice phenological characteristics to identify the key growth period of Rice--transplanting period.Combined with the rice planting division and the rice lunar calendar information of each region,the effects of permanent water body,evergreen vegetation and snow and other ground types are eliminated,so as to realize the rapid extraction of rice planting spatial distribution and planting area in the study area.Compared with the statistical yearbook data,the error of rice planting area in 2015 and 2016 in the research area obtained by the research of rice planting extraction method based on time series MODIS data is about-15.81% to 25.75%,and the spatial distribution is more consistent with the actual situation in the research area.(2)The spatial distribution map of rice planting and the extraction of rice planting area from 2004 to 2018 in Cengong County,Southeast Miao and Dong Autonomous Prefecture of Guizhou Province Based on long-time series Landsat data are realized.In this paper,CCDC(continuous change detection and classification Classification)algorithm is combined with time series model,which is applied to the remote sensing data Landsat of optical satellite in the cloudy and foggy area of our country.The image data set without cloud and strip for many years in this research area is obtained successfully.Then the rice recognition threshold model is established by combining the rice phenology characteristics and the ground measurement data,and the rice species with high accuracy in the small area is realized Plant spatial distribution and plant area extraction.The above results show that the validation accuracy of the field data is about 87%.Compared with the statistical yearbook data,the error of rice planting area obtained for many years is about-5.2% to +8.87%.(3)The rice planting area extraction based on the multi-source data(Landsat data,sentinel series satellite data)of RF(random forest)classification method is realized,taking Hainan Province as an example.In this paper,using the online cloud computing platform provided by Google to call the Landsat and sentinel data in the database,and then using the current mainstream methods to coordinate different satellite data in the same location,using the method of monthly synthesis based on the longitudinal number(mode)and the method of random forest classifier combined with visual interpretation to obtain the sample point data of rice planting in Hainan Province successfully Spatial distribution map and planting area.The overall accuracy is 88.01% and kappa coefficient is 75.01%,which can meet the practical application requirements.