Research On Remote Sensing Image Recognition And Monitoring Of Rice Planting Area In Yunnan-guizhou Plateau Based On Google Earth Engine

In the remote sensing image recognition of plateau and mountainous areas,the fragmentation of land blocks and the complexity of land use types bring great difficulties to image recognition.This thesis uses Google Earth Engine as the data processing platform,Qujing city in Yunnan province as the research area,and Landsat series data as remote sensing data.A series of vegetation indices are calculated.According to the change characteristics of these vegetation index time series,the key vegetation index for rice extraction is determined,and then different classification methods are used to extract the rice planting information in the study area.In this thesis,four different classifiers are selected for training and classification,which are Classification and Regression Tree,Support Vector Machine,Maximum Entropy Model and Random Forest.The sample points collected by Google Earth Pro visual interpretation of historical images were used to verify the experimental results.Finally,the best classification method was selected to identify the rice planting information of Qujing City in the past 30years,and the rice planting information was monitored by remote sensing and the driving force was analyzed.The main research contents and conclusions of this thesis are as follows:(1)Remote sensing feature extraction of rice based on vegetation index time series analysis.Based on online selection of sample points in the study area in the GEE data,calculating a series of vegetation index,based on time series curve of each vegetation index time series analysis,selection significantly differentiate between across the study area of vegetation index,Harmonic Analysis of Time Series is used to smooth the time series curve,according to the time sequence curve after the smooth,selects the Enhanced Vegetation Index,Land Surface Water Index,Normalized Difference Built-up Index and Modified Normalized Difference Water Index are used as the vegetation index for the extraction of rice planting information.(2)Extraction of rice planting information in Qujing based on multi-classification method optimization.In GEE,the rice planting information of nearly 30 years in the study area was extracted by using random forest classification method according to the constructed features.The rice planting information of 1990,1995,2000,2005,2010,2015 and 2019 in the study area was obtained.The total area was 517.47km2,529.71km2,446.18km~2,549.25km~2,549.90km~2,593.69km~2 and614.03km~2,respectively.The overall accuracy reached 79.70%,81.10%,84.33%,89.05%,89.53%,91.25%,94.45%,respectively.(3)Remote sensing monitoring and driving factors analysis of rice planting area based on spatial analysis.Based on Pearson correlation analysis,spatial auto-correlation tests,center of gravity analysis,spatial variation analysis on spatial analysis technologies such as the time and space changes in 7 years rice planting information analysis and comparison,the area was revealed nearly 30 years change trend and the influence factors of the rice planting information for rice yield estimation in the study area,the condition of production monitoring and rice planting of government departments to formulate relevant policies to provide certain scientific basis.