An improved multi-temporal masking classification method for winter wheat identification

Abstract

In the field of crop identification with remote sensing technology, current multi-temporal methods usually do not made full use of target crop's temporal features and spectral features. An improved multitemporal masking classification method was proposed for winter wheat identification in Jiaodong Peninsula. The improved method using four temporal MODIS NDVI product images and two temporal TM surface reflectance imagines could better conduct both temporal feature recognition and spectral feature recognition. First, a winter wheat mask was generated from four proper temporal MODIS NDVI product images to distinguish winter wheat from the other local crops; second, the winter wheat mask was applied to a multi-time phases combined TM image. Then according to classes' spectral separability, a set of TM bands were selected to form spectral space for classification, and winter wheat was identified by spectral classification in the TM spectral space. In the study area, the identification accuracy reaches 94.92%. The results indicate that the appropriate winter wheat mask design and spectral classification bands selection could help the improved multi-temporal method to get high winter wheat identification accuracy.