Research on Extraction Method of Rocky Desertification Information Based on Landsat——Take Chenxi County as an Example
[Purposes]The author aims to study and analyze the extraction method of rocky desertification in Chenxi County,Huaihua City,Hunan Province,in order to provide a reliable basis for ecological resto-ration of southern hills.[Methods]Based on Landsat-9 OLI image,combined with the exposure rate of bedrock and vegetation coverage,the rocky desertification classification index was constructed,and the expert experience decision tree and object-oriented model based on vector machine were used to extract rocky desertification,and the degree of rocky desertification was divided into five levels:severe,moder-ate,mild,potential and no rocky desertification.The accuracy was compared by true color remote sens-ing image,Google Earth historical image and field check.[Findings]The total validation accuracy of ex-pert experience decision tree method(89%)was higher than that of vector machine-based object-oriented method(64%),with Kappa coefficients of 0.852 052 and 0.530 209,respectively.The rocky de-sertification was generally mild and mainly distributed in the northeastern part of the study area.[Con-clusions]The study suggests that the decision tree extraction method based on expert experience should be widely applied in the hilly areas of South China to improve the accuracy and practicability of rocky de-sertification monitoring,and the characteristic areas such as steep terrain and sloping farmland should be paid more attention.
NETL
NSTL
万方数据
