www.mdpi.com/journal/remotesensing

ABSTRACT: Crop classification plays an important role in effective and controllable agricultural management. In this paper, nine scenes of Landsat8_OLI data in 2013 were collected for cotton planting area extraction in Shawan country of Xinjiang Uygur Autonomous Region, China. Normalized Difference Vegetation Index (NDVI) time series were generated to characterize the phenological pattern of each crop type. The optimal temporal reflectance image was chosen by analyzing the difference of NDVI profile between cotton and other crop types. The hierarchical classification strategy was performed on the three features of NDVI time series, NDVI statistics and reflectance. Firstly a simple decision tree was built on NDVI statistics and reflectance to extract vegetation cover; then various types of crops were distinguished by support vector machine (SVM) and maximum likelihood supervised classifier (MLC), thereby cotton plating area was extracted. A comprehensive evaluation for the cotton extraction map was performed using human-computer interaction visual interpretation and ground truth data. Results showed that MLC achieved the accuracy of 97.56 % for cotton extraction. The cotton extraction map was very consistent with the ground truth data. This multi-temporal classification method is promising for crop extraction even for land cover classification. 1.

30 de agosto de 2004.

CALMIT wishes to thank the Nebraska Environmental Trust and the Cooperative Hydrology Study Sponsors for their funding of this research. Other thanks go to the many Natural Resources Districts (NRDs) for their help in providing field and ancillary data to supplement the land cover classification including: Central Platte NRD, North