Vegetation Fractional Cover Extraction and Spatiotemporal Variation Analysis Based on Improved Normalized Difference Vegetation Index(NDVI)and Dry Fuel Index(DFI)Model
[Objective]By introducing the DBSCAN clustering algorithm and QuickHull convex hull detection algorithm,an adaptive endmember eigenvalue extraction(AEEE)algorithm is proposed to address the issue of manual selection of feature pixel candidate areas in the pixel ternary model combined with the pure pixel index-2D scatter plot(PPI-2DSP)algorithm.The AEEE algorithm is utilized to assess the photosynthetic vegetation coverage(fPV),non-photosynthetic vegetation coverage(fNPV),and bare soil coverage(fBS)in Shenmu city and analyze their spatiotemporal variations.The effectiveness of the algorithm is verified,providing reference for the evaluation of the ecological environment and the study of vegetation fractional cover change patterns in the region.[Method]Using Landsat series satellite remote sensing images as the data source,the remote sensing data is preprocessed first.Then,the normalized difference vegetation index(NDVI)and dry fuel index(DFI)of pixels are calculated.Feature pixel candidate areas are obtained through the following 4 steps:1)Reducing the data volume using a random sampling module;2)Clustering with the DBSCAN algorithm to remove outlier data and identify the largest cluster;3)Computing the convex hull using the QuickHull algorithm to construct the boundaries of feature triangles;4)Calculating the vertices of the largest area triangle formed by three points within the convex hull point set.Regions within a specified range centered at the three vertices are selected as feature pixel candidate areas with a vertex threshold(θ).After reducing the computational complexity through the Minimum Noise Fraction transformation of preprocessed images,the pure pixel index is calculated using the PPI algorithm.Pure pixels with a pure pixel index greater than 5 are extracted using feature pixel candidate areas.The arithmetic mean values of the NDVI and DFI for these pixels are used as endmember eigenvalues into the pixel ternary model to calculate fPV,fNPV,and fBS,followed by an analysis of their spatiotemporal variations.[Result]The eigenvalues calculated by the AEEE algorithm for Shenmu city from 2000 to 2022 are close to those selected by the PPI-2DSP algorithm,with an average relative error of approximately 7.35％.When applied to the NDVI-DFI model,the estimation of fPV and fNPV for Shenmu city using AEEE exhibits errors of 4.79％and 5.05％,respectively,compared to the traditional method,meeting accuracy requirements.On a temporal scale,from 2000 to 2022,fPVan fNPV in Shenmu city showed an overall fluctuating growth trend,increasing at average annual rates of 0.52％and 0.22％,respectively.On a spatial scale,from 2000 to 2022,fPV and fNPV in Shenmu city exhibited a trend of rapid growth in the southeast and slower growth in the northwest,with fPV primarily characterized by two main change intensities:increase(39.8％)and basically unchanged(28.0％).[Conclusion]The AEEE algorithm is suitable for the adaptive extraction of endmember eigenvalues for photosynthetic vegetation(PV),non-photosynthetic vegetation(NPV),and bare soil(BS).It addresses the issue of the PPI-2DSP algorithm relying on manual selection of feature pixel candidate regions.
normalized difference vegetation index(NDVI)and dry fuel index(DFI)modelvegetation fractional coverLandsatadaptive endmember eigenvalue extraction(AEEE)algorithmShenmu city of Shaanxi Province
NETL
NSTL
万方数据
