Extraction method for farmland from the plain-mountain region by remote sensing in farmland fertility survey

The accurate acquisition of the cultivated land resource information is important in the farmland fertility survey and evaluation. Aiming at the mixed spectral characteristics of the plain-mountain region，Qingzhou county of Shandong province was taken as a case to extract the farmland resources. Based on the experiment of the spectral knowledge classification and the supervised classification, an extraction method for farmland by remote sensing was proposed. On the basis of the accuracy assessment of the supervised classification, the leak pixels were added to the type and the wrong pixels were removed from the type to determine the pixels of each farmland type contained, Then, combing with the spectral characteristics, experience, land utilization condition information, morphology, cultivated land productivity survey data and other background information, the extracted rules of each cultivated land type was established. The results showed that the classification accuracy of this method was high, which could reach up to 88.57%, and the automatic and fast classification with high accuracy of the region was realized. This method can be a reference for the regions with the similar situation.