Farmland shelterbelt is a major component of land reclamation, farmland protection and the engineering construction of ecological environmental protection. And the information acquisition of farmland shelterbelt is an important approach to supervising engineering, such as land reclamation and so on. The automatic recognition method of farmland shelterbelts was explored by taking full advantages of the spectral and morphological features in high spatial resolution remote sensing data based on GeoEye-1 satellite remote sensing image (0.5m-resolution). Firstly, the classification decision tree was established by making use of the normalized difference vegetation index (NDVI) and two-dimensional entropy of GeoEye-1 satellite remote sensing image. Secondly，the preliminary extraction results of farmland shelterbelts were found out by combining with the auxiliary data. Thirdly, the preliminary extraction results were processed by morphology operations, including expansion, hole filling, de-noising and thinning, to get the continuous and refined extraction results. The partial region of the oasis in the southern margin of Badain Jaran Desert in Linze County, Gansu Province was taken as a typical study area for instance validation, which is located in the central Hexi Corridor. The experimental result indicated that by using the method the automatic recognition precision of farmland shelterbelts were all above 92% and the average accuracy reached 92.97%, the spatial coincidences were all above 86% and the average anastomosis reached 93.13%. So it can satisfy the actual demand in supervising engineering, such as land reclamation and so on. The method can provide scientific support for the farmland shelterbelts construction and the related engineering supervision.