Study Of Remote Sensing Monitoring Technology For Returning Farmland To Wetland Based On GF Data

The"Returning Farmland to Wetland"project is an important measure for my country to protect wetland natural resources and restore the wetland ecosystem.However,the project of returning farmland to wetland is a long process.Due to the complex environment of the wetland ecosystem and poor accessibility,traditional wetland survey methods cannot meet the rapid monitoring requirements for project implementation.In response to the above problems,this research aims to analyze the characteristics of the areas where the conversion project has been implemented,and find the remote sensing information extraction technology to accurately identify the conversion land parcels,and provide technical support for my country’s scientific conversion of farmland and restoration of the ecological functions of wetlands.This research uses the Heilongjiang Sanjiang National Nature Reserve(hereinafter referred to as Sanjiang Reserve)as the research area of the project of returning farmland to wetland,and uses domestic high-resolution satellite image data to explore the remote sensing extraction method of land use type information suitable for the characteristics of the research area.Based on the extraction results,the spatial distribution characteristics of converted farmland to wetland and the restoration status of wetland in the study area are analyzed.At the same time,in view of the advantages of GF-6 WFV data in vegetation information monitoring,the best band combination of GF-6 WFV band for the identification of vegetation types in the study area is sought.The main con-tent and conclusions are as follows:(1)Based on the GF-6 WFV image,from the perspective of image information characteristics and spectral separability,using information index,correlation coefficient,OIF index,spectral characteristic curve statistics,JM distance and other methods,we comprehensive select the best band combination for vegetation type identification.It was finally determined that under the band combination 4(R)-5(G)-1(B),the vegetation information is the most abundant,which is the best band combination we need.(2)Based on the sample data,this research constructs a new feature selection method—RF-RF RFE algorithm.The algorithm is based on the random forest algorithm and packaging ideas.It selects important feature sets for classification from spectral features,exponential features,texture features,geometric features,location features,and interclass correlation features.Compared with a single feature selection algorithm,the optimized RF-RF RFE algorithm can not only filter features according to their importance,and improve the calculation efficiency,but also can take the feature set as the research object and directly give the optimal feature set.Secondly,based on the optimal feature set,comparing the three classification methods of CART decision tree,SVM model and random forest,it is found that the accuracy of random forest classification is the highest.Finally,based on the analysis of the importance of features,the results show that multiple types of features help improve classification accuracy.Among them spectral features,index features,and interclass distance features among the interclass correlation features are important features of the Sanjiang Nature Reserve.(3)The RF-RF RFE feature selection method and the random forest classification method are used to complete the land type information extraction of the Sanjiang Nature Reserve.The spatial analysis method is used to obtain the spatiotemporal dynamic change data of the protection area before and after the conversion of farmland to summarize the results of the conversion of farmland to wet projects and the types of regions.The law of temporal and spatial changes of area and the reasons for the changes.The results showed that the area and distribution of various types of land changed greatly before and after returning farmland to wetland.The area of cultivated land decreased by 293.53km~2,and the area of water and peatland increased by 238.86km~2and 130.25km~2,respectively.Among them,the area of wetlands in the northwestern part of the Amur River Basin Reserve has increased significantly,and the area of watersand herbaceous swamps have increased significantly.The increased peatlands are mainly located in the Wusuli River Basin Reserve,which is mainly transformed from herbaceous swamps.The scope of implementation of the project of returning farmland to wetland during 2016-2019 is located in the core area of the study area.According to statistics,the area of cultivated land and herbaceous swamp in the core area has decreased significantly,while the area of water and peatland has increased,mainly due to the following reasons.The following two points,one is the implementation of the project of returning farmland to wetland in the study area,which has reduced the area of cultivated land in the core area;the second is that the heavy rainfall in 2019 has accelerated the transformation between local types.(4)Based on the land use information before and after returning farmland to wetland,we extract the results of the spatial and temporal changes of landscape ecological risk in the study area and analyze the reasons for the changes.The results show that the area of ecological risk levels in the study area changes to a certain extent during the period before and after the conversion of farmland.The early period of conversion is mainly low,medium and high risks,and the latter period is mainly low,low and medium risks.The risk of the overall landscape ecology shows a downward trend.Through the spatial distribution,it can be found that the landscape ecological risk pattern of the study area also has obvious changes.Among them,the landscape ecological risk of the core area of the study area shows a downward trend,which is mainly due to the implementation of the project of returning farmland to wetland,which reduces the cultivated land area and the fragmentation of wetland landscape in the core area.In addition,the middle risk area in the northern part of Amur Basin has a larger area transformed into a lower risk area,which is mainly due to the heavy rainfall in 2019,which makes the cultivated land and other types in this area submerged by water,which makes the landuse type single and reduces the landscape ecological risk of this area to a certain extent.