Comparison of vegetation coverage models based on comprehensive segmentation advantages

On the basis of Landsat 8 OLI remote sensing images, nine vegetation indices were selected to analyze the correlation between vegetation indices and field-measured vegetation coverage. A sensitive vegetation index was selected based on the segmentation method, and regression and FCD (forest canopy density mapping model) models were used to invert the vegetation coverage in the Weigan-Kuqa River delta oasis. The following results were obtained. 1) Using the measurement data, the change proportion of different vegetation indices in different vegetation coverage ranges was calculated, and 0.3 and 0.7 were determined as the segmentation points of vegetation coverage in the study area. 2) The modeling accuracy of the segmented regression and FCD segmentation models was approximately 79%; however, the R² (0.832) and RMSE (0.154) values of the FCD segmentation model were higher and smaller, respectively, than those of the segmented regression model. Thus, the FCD segmentation model was established as providing better validation results (PRECISION of 82.018%). The FCD segmentation model is therefore considered more suitable for inversion of the total vegetation coverage in the study area, and will contribute to the quantitative monitoring of vegetation coverage and evaluation of the ecological environment in arid areas.