[Objective] The extraction of vegetation cover and optimization of vegetation indices in desert hinterland oasis were analyzed and studied in order to providing scientific basis for select the optimal vegetation indices to invert the vegetation cover status of extreme arid zone oasis. [Methods] Natural vegetation cover data from the Dariyabui Oasis in the hinterland of the Taklamakan Desert were obtained from UAV aerial photography sample images and used as the benchmark. A variety of typical vegetation indices were extracted from Sentinel-2B satellite images, and a vegetation index-vegetation cover statistical model was established using regression statistics to determine the optimal vegetation index for inversion to quantify arid oasis vegetation cover at the satellite image element scale. [Results] ① The accuracy of vegetation cover of the extracted samples using Image J software was high, and the overall accuracy reached 88.67%. ② The soil-regulated vegetation indices (SAVI, MSAVI) performed well as shown by standard regression coefficients and coefficients of determination, and had good applicability in reflecting natural vegetation cover changes in extreme arid zones. [Conclusion] In an extreme arid zone, Image J software did well in extracting sparse vegetation cover, and SAVI and MSAVI are the more suitable vegetation indices for oasis vegetation cover change studies.