Abstract: In recent years, extreme rainstorms and natural disasters have occurred frequently, which causing severe damage to farmland．A method of identifying hump and concave features of farmland using Farmland Digital Terrain Model（FDTM）was proposed．First, aerial image of experimental plot was handle based on SfM（Structure from Motion）technology and high precision farmland FDTM was obtained．Variation trend of elevation variance of FDTM with local window scale was analyzed, and scale interval of analysis window as 31×31 to 51×51 pixels was determined．Second, 315 elevation extremes were extracted under local window of 51×51 pixels, which were further evaluated by using 3 topographical factors of elevation, topographic relief and slope to obtain multi-windows comprehensive membership degrees．Finally, threshold value of multi-windows comprehensive membership degrees was determined as 0.627 according to Sturgis method．16 elevation extremes were confirmed as hump features’ vertices, and outline of hump features were obtained by combining with contour map．Similarly, a Reverse Farm Digital Terrain Model（RFDTM）was developed to transform concave features into hump features, and 9 farmland concave features were identified．Research results could provide theoretical basis and methodological support for farmland reclamation and precise land leveling operations．