Information Extraction And Dynamic Monitoring Of Rice Planting In Shenyang City Based On Landsat8 Remote Sensing Image

The total output of rice in China ranks first in the world,planting area is the second largest in the world,rice has become one of the most important grain crops in our country.Timely understanding of rice growth and accurate mastery of rice planting area,timely understanding of rice growth,accurate control of rice planting area and accurate estimation of yield are important for rural development and government's policy of agricultural production,and ensuring food safety in China and even in the world.However,because of the influence of many factors,the application of traditional visual interpretation cannot meet the requirements of precision.With the updating of remote sensing data and the development of technology,the object oriented classification better classification effect and higher classification accuracy.This paper takes Shenyang city of Liaoning Province as the research area,the Landsat8 OLI remote sensing images of 9 different phases from 2013 to 2016 are the basic data,ENVI5.4 software was applied to extract rice planting information in the study area based on the methods of unsupervised classification,supervised classification and object oriented classification,and the dynamic change monitoring of rice was done by the method of classification and then comparison.The main contents and results of this paper are as follows:(1)Preprocessing of remote sensing images in the study area.First,the radiometric calibration of the multispectral and panchromatic bands is carried out respectively,the DN value is converted to the earth surface ground surface reflectance,surface temperature and other parameters.Then the multi spectral image and the panchromatic band image after the atmospheric correction are fused,a new NNDiffuse Pan Sharpening fusion method is used to improve the spatial resolution of the image and preserve the color,texture and spectral information well.Finally,the research area is cut,image mosaic and image mosaic to get the image data of the study area through the vector boundary data of the research area.(2)Remotely sensed image separability analysis.The growth period data from May to October by rice,transplanting stage,booting stage,heading stage and milk stage and mature stage as the main growth period of rice yield estimation by remote sensing.A 16 level historical satellite image with a maximum pixel resolution of up to 1.78m as a real region of interest validation sample,the study of feature points Landsat8 data selected as training samples,and the terrain,the northeast,the phenological data of elevation data,vegetation coverage(NDVI index)as auxiliary data,extraction of water,land,buildings,rice,corn,6 kinds of key features,separability analysis and spectrum analysis and the key features of the separation,obtained the highest one image.(3)Information extraction of rice in Shenyang.Two kinds of methods based on pixel(unsupervised classification,supervised classification)and object based(rule based object oriented,sample based object oriented)were applied to rice classification.Then,the classification principle,classification process and classification results were analyzed.Finally,the classification accuracy of the different classification methods was compared through the confusion matrix,and the best classification method was obtained,and the rice planting information was extracted.(4)Comparison of estimation area and statistical data of rice by remote sensing.Using the Statistic tool in ENVI,the total number of rice in Shenyang was calculated,and the rice planting area in 4 years was counted,and the accuracy of rice area estimation was analyzed by comparing the data with the data of Shenyang Statistical Bureau.(5)Study on the dynamic monitoring of rice planting area.In 2013 and 2016,the two phases of the image after the object-oriented processing are monitored and studied dynamically.In this paper,the dynamic changes of rice comparative law comparison method and classification,finally obtains the research area of rice crop distribution,rice area,in the past 4 years,the statistical table of rice planting area change,land use transfer map and land use transfer matrix,and the dynamic change of rice planting was obtained from pixel,percentage and area.By comparing the spatial accuracy of different classification methods,it is concluded that the accuracy of object classification is the highest,the total classification accuracy is 95.4%,and the Kappa coefficient is 0.94.The method was applied to the statistics of rice planting area in recent 4 years.It was concluded that the estimated results of remote sensing in 2016 were the closest to the true value released by Statistics Bureau,reaching over 91.60%.