UAV-based hyperspectral analysis and spectral indices constructing for quantitatively monitoring leaf nitrogen content of winter wheat.

In flag leaf and flowering stages of winter wheat, unmanned aerial vehicle (UAV)-based and ground-measured hyperspectral data were collected simultaneously, and leaf nitrogen content (LNC) data were then measured in a laboratory. First, the accuracy of UAV-based hyperspectral data was analyzed using ground-measured hyperspectral data, and the analysis showed that the effectiveness and spectrum sampling precision of the UAV-based hyperspectral data are reliable. Hyperspectral characteristic analysis of winter wheat canopies of different LNCs was also conducted. Second, representative spectrum bands that are sensitive to the LNC of winter wheat were extracted through first-order differential spectral, continuum-removed reflectance, and band correlation prediction threshold methods. The optimal band combination that is sensitive to the LNC of winter wheat was obtained by comparing and analyzing the representative spectrum band results. Thus, several LNC spectral indices (LNCSI) were established through ratio, difference, and normalization methods, and linear regression statistical models for quantitatively simulating LNCs were established using the LNCSIs. Finally, comprehensive and comparative analyses of the LNCSIs and the inversion values of the LNC using the LNCSIs confirmed that the LNCSIs are effective in quantitatively inversing the LNC of winter wheat.