| Remote sensing technology has the advantages of high information quantity,macro view,objectivity and timeliness.It is an important technical means for agriculture,forestry,land resources and other departments to monitor and manage.The accurate extraction of crop planting area is closely related to crop growth monitoring,yield prediction and market price.The traditional methods of crop area extraction mostly rely on satellite data such as MODIS,TM,SPOT,so it is difficult to achieve the balance between time and space,and the precision of extraction still needs to be improved.In recent years,with the introduction of various new sensors,the interaction between different data has become a hot topic.Therefore,how to use multi-source remote sensing data to construct high spatial resolution time series will have certain application value.In order to solve the problem of poor continuity of single source remote sensing data with high spatial resolution in ground object recognition,this study takes GF-1 and Landsat-8 images as the basic data sources,and puts forward a joint construction of vegetation index time series data with high spatial and high time resolution.A series of processing processes,such as data selection rules,vegetation index conversion and so on,are used to reduce the data differences caused by different data acquisition platforms.In order to evaluate the effect and applicability of GF-1 and landsat-8 vegetation index time series,this paper selects Shijin irrigation area as the experimental area,and compares the clustering accuracy of vegetation index time series data with high spatial and high time resolution through five commonly used distance measure methods.The best data source and identification method of planting structure extraction of Shijin irrigation area are determined by comparing with the traditional classification method based on MODIS,TM as remote sensing data source.In addition,based on the spatial distribution of planting structure in Shijin irrigation area,the grain yield is estimated with spectral information,precipitation and temperature as influence factors.The experimental results show that:(1)The relationship between Landsat-8 and GF-1 NDVI data is highly linear,and the relationship between GF-1 NDVI fitting data and Landsat-8 NDVI is enhanced after conversion equation,which effectively reduces the degree of data difference.(2)The vegetation index time series based on multi-source data can give full play to the respective advantages of GF-1 and Landsat-8 images.Compared with its single source data,it has a higher classification accuracy in ground object recognition.The classification results of the five distance measures are all of high accuracy,and the overall classification accuracy is as high as 96.09%.(3)The overall classification accuracy of MODIS NDVI time series is 84.13% and the coefficient of Kappa is 0.79.And the overall classification accuracy of multi-temporal Landsat-8 is between 76% and 86%,and the coefficient of kappa is between 0.6 and 0.75.Compared with the classification results of multi-source time series interpolation data,the traditional classification method has lower extraction accuracy,which further verifies the feasibility of remote sensing recognition by GF-1 and Landsat-8 time series interpolation data.(4)Based on the spatial distribution of the planting structure in Shijin irrigation area,the spectral yield model,the meteorological yield model and the comprehensive spectral meteorological yield model are established respectively.Compared with the spectral yield model and the meteorological yield model,the correlation coefficient of the comprehensive spectral meteorological yield model is as high as 0.89,and the decision coefficient of a cross-test is 0.79,which is generally superior to the other two yield estimation models.The study provides a new technical tool for the development of remote sensing identification and estimation of grain production in small and medium scale regions,and provides guidance for governments and other relevant departments to understand food production trends in different ecological regions in a timely manner.It also provides a reference for the formulation of grain trade and macro-control policies. |
