The abandoned cropland is spreading widely all over the world. Traditional survey methods can not meet the need of timely and accurate grasp of abandoned cropland. In view of the insufficiency of optical remote sensing images, and that it is difficult to accurately identify and extract the abandoned cropland in karst rocky desertification land from the serious mixed pixels, taking Xifeng County of Guizhou Province as an example, the high-resolution image and rocky desertification grade data were used to identify and extract different rocky desertification grade cultivated land plots. The time series NDVI data of 2003-2018 obtained by calculating the Landsat data by MVC method was superimposed, and the spatial and temporal distribution of the abandoned cropland and reclamation in Xifeng County in 2003-2018 was extracted. The results showed:(1) The combination of cultivated land plot and time series NDVI could accurately identify and extract abandoned cropland. This method had a good application in the area lacking of optical data and the cloudy rainy mountain area where the cultivated land is broken. (2) In 2003-2018, the abandoned cropland in Xifeng County was randomly scattered and mainly distributed in the potential, mild and moderate rocky desertification area in the north and east. In 2008, the largest abandoned area was 2 545.6 hm² and the highest rate of abandonment was 7.7%. In the period of 2003-2018, there were 7 plots of abandoned land that lasted for a maximum of 12 years, and 8 plots accumulated for up to 15 years. (3) In 2003-2018, the reclamation rate of Xifeng County was basically consistent with the fluctuation trend of the abandonment rate, but the response to the reclamation rate was delayed by one year. There was a certain difference between the non-karst cultivated land reclamation behavior and the rocky desertification cultivated land reclamation behavior. The higher rock desertification grade negatively affected the reclamation behavior. The research results could provide an efficient and feasible idea and method for the accurate identification and extraction of abandoned cropland in karst cloudy and rainy mountainous areas, also provide the accurate and basic data for driving factor analysis, trend monitoring and risk assessment, effect assessment and policy design in abandoned cropland.