Desertification Information Extraction And Change Detection In Gentle Slope Area Of Yanchi County

In order to obtain the dynamic changes of seven land use cover types in cultivated land,grassland,woodland,bare land,sandy land,water body and construction land in the study area,this thesis uses 25 features of spectrum,texture and topography as classification feature variables.Object segmentation,using the random forest(RF)classification algorithm to classify the Landsat TM and the Landsat OLI remote sensing images of the two phases of the study area in 2010 and 2017,and establish a confusion matrix for the classification results to make classification accuracy evaluation,and test the classifier the overall classification accuracy is 94.95%,and the Kappa coefficient is 0.94,which indicates that the classification algorithm of object-oriented combined random forest can be effectively applied to the classification of features in desertification areas.Through the evaluation of the feature importance of the random forest classification model,it is found that Landsat OLI B1 band,PC1 mean feature,normalized difference vegetation index(NDVI),BSWIRI band and the green feature of cap change have great influence on the classification accuracy of the model.Characteristic variables such as the characteristic band.When the land use coverage change is detected,the total area coverage of the various types of land in the classification results shows that the sand area decreased from 461.84 km2 in 2010 to 206,76 km2 in 2017,a total reduction of 255.08 km2.The main conversion targets are bare land and grassland,and the corresponding conversion areas are 129km2 and 41.26km2 respectively,which reflects the treatment effect of the research area on desertification key land types in recent years.The coverage area of another type of desertification key land types increased from 456.27km2 in 2010 to 852.42km2 in 2017,with a total increase of 396.15km2.The main source of growth was the conversion of sandy land and grassland types into bare land.The NDVI-Albedo feature space is constructed to determine the quantitative relationship between NDVI and Albedo,and the desertification difference index(DDI)of the study area is calculated according to the research conclusions of Verstrate and Pinty,and the Jenks natural discontinuity classification method is used.The Desertification Difference Index(DDI)is divided into four levels:severe desertification,moderate desertification,mild desertification and non-desertification.By measuring the change of desertification area in two periods,the coverage areas of non-desertification and severe desertification in the study area are reduced by 105.25km2 and 83.16km2,respectively,which is-15.59%compared with the non-desertification and severe desertification areas in 2010,respectively.And-16.88%,its coverage area showed a decreasing trend,which was consistent with the trend of reducing the grassland land type by 605.8km2and the sand cover area by 255.08km2.At the same time,the coverage areas of mild desertification and moderate desertification increased by 113.94km2 and 74.37km2 respectively,which is consistent with the trend of 319.15km2 increase in bare land area in the classification results of ground objects.To some extent,the use of NDVI-Albedo feature space is illustrated.The method of retrieving the desertification difference index and classifying the desertification index by the natural discontinuity classification method is applicable to the desertification information extraction in this study area.