Object oriented extraction of reserve resources area for cultivated land using RapidEye image data

Identify the amount and spatial distribution of reserve cultivatable land resources is the basis for its development to increase crop planting areas. Taking Jiaxiang county of Shandong Province of China as a case study, this paper conducted image segmentation and merge based on RapidEye image data (5m spatial resolution) after data preprocessing. Then, object-oriented approach was used to classify land use information and the reserve resources of arable land were extracted from them. The results showed: (1) 30% and 80% of the scale level and merge level for image segmentation and merge were chosen for getting better results of independent polygon division based on object-oriented approach; (2) Comparing with K near value method (KNN) and principal component analysis method (PCA), support vector machine (SVM) method had 78% of the highest overall accuracy for the supervised classification; (3) The overall land use classification accuracy was 90.4% verified by field survey data and 1:10000 land use map in 2011, Kappa coefficient was 0.8784. Therefore, using high spatial resolution image can improve the classification accuracy for the reserve cultivatable land resources; (4) Bare land, wild grassland, mudflats and reed land were the main reserve resources of cultivated land for the study region. The area was 2640 ha occupying 2.95% of the total land area only. The area of bare land and wild grassland accounts for 61% and 35% of the reserve cultivatable land resources. Thin soil thickness and lack of irrigation facilities were major limit factors for its development to cropland.