Abstract: Abstract: Farmland is an important agricultural resource, which is directly related to the food security and is concerned by the public and government. As an important method, remote sensing images are usually interpreted to acquire the farmland area. However, for the reasons of spatial resolution of remote sensing images, many small features and linear features are interpreted into farmland, which surely affect the final farmland area. Thus, subtracting the small features and linear features is a key method to improve the accuracy of farmland area during the process of remote sensing survey. In this study, the Suixi County of Anhui province was selected as the study area. In addition, the GF-1 image with 16 meter resolution was used as data source. In order to get exact farmland area, different methods were used in various steps. Firstly, the farmland was interpreted by the fusion of image texture and spectral information according to the GF-1 image. Secondly, linear features were subtracted by remote sensing interpretation and spatial calculation of Geological Information System (GIS), and small features were subtracted according to the 36 sampling frames. Thirdly, farmland areas of interpretation in Suixi County were accurately calculated by subtracting the small features and linear features. Four important conclusions were got in this study, which were list as follows. (1) The total farmland area in the Suixi County was 1320.64km2 according to the results of interpretation. The farmland mainly distributed in the central regions (such as Sipu Town, Baishan Town) and southern regions (such as Shuangduiji Town, Nanping Town), while it relatively distributed small in the northern regions (such as Suixi Town, Liuqiao Town). (2) Liner features in the Suixi County mainly contains road, river and channel. There were 45.14km2 liner features which were interpreted into farmland according to the GF-1 image. The area of these liner features took up 3.42% of the total farmland area in the Suixi County. (3) According to the calculation of 36 sampling frames, the subtracting coefficients of small features were between 1.27% and 7.92%, and the mean of subtracting coefficient for the small features was 3.51% in the Suixi County. In addition to the linear features, there were still 44.77km2 small features which were interpreted into farmland. (4) Based on the results of interpretation, the finial farmland area in the Suixi County was 1230.73km2 after subtracting the small features and linear features. The absolute value of mean error for the farmland area was from 6.57% before subtracting the small features and linear features, while it was only 0.68% after the subtraction in the Suixi County. As for the 36 sampling frames, the absolute value of mean error was from 1.77% before the subtraction to 1.91% after the subtraction, which was not significant. The results in this study implicated many small features and linear features were interpreted into farmland. Subtracting the small features and linear features could obviously improve the accuracy of interpreted farmland area. This study could be helpful for improving accuracy of interpreted farmland area during the process of remote sensing survey.