Yu-Jie Wang1, Tie-Han Li1, Ge Jin1, Yu-Ming Wei1, Lu-Qing Li1, Yusef K Kalkhajeh2, Jing-Ming Ning1, Zheng-Zhu Zhang1. 1. State Key Laboratory of Tea Plant Biology and Utilization, Anhui Agricultural University, Hefei, China. 2. Anhui Province Key Laboratory of Farmland Ecological Conservation and Pollution Prevention, School of Resources and Environment, Anhui Agricultural University, Hefei, China.
Abstract
BACKGROUND: Rapid and accurate diagnosis of nitrogen (N) status in field crops is of great significance for site-specific N fertilizer management. This study aimed to evaluate the potential of hyperspectral imaging coupled with chemometrics for the qualitative and quantitative diagnosis of N status in tea plants under field conditions. RESULTS: Hyperspectral data from mature leaves of tea plants with different N application rates were preprocessed by standard normal variate (SNV). Partial least squares discriminative analysis (PLS-DA) and least squares-support vector machines (LS-SVM) were used for the classification of different N status. Furthermore, partial least squares regression (PLSR) was used for the prediction of N content. The results showed that the LS-SVM model yielded better performance with correct classification rates of 82% and 92% in prediction sets for the diagnosis of different N application rates and N status, respectively. The PLSR model for leaf N content (LNC) showed excellent performance, with correlation coefficients of 0.924, root mean square error of 0.209, and residual predictive deviation of 2.686 in the prediction set. In addition, the important wavebands of the PLSR model were interpreted based on regression coefficients. CONCLUSION: Overall, our results suggest that the hyperspectral imaging technique can be an effective and accurate tool for qualitative and quantitative diagnosis of N status in tea plants.
BACKGROUND: Rapid and accurate diagnosis of nitrogen (N) status in field crops is of great significance for site-specific N fertilizer management. This study aimed to evaluate the potential of hyperspectral imaging coupled with chemometrics for the qualitative and quantitative diagnosis of N status in tea plants under field conditions. RESULTS: Hyperspectral data from mature leaves of tea plants with different N application rates were preprocessed by standard normal variate (SNV). Partial least squares discriminative analysis (PLS-DA) and least squares-support vector machines (LS-SVM) were used for the classification of different N status. Furthermore, partial least squares regression (PLSR) was used for the prediction of N content. The results showed that the LS-SVM model yielded better performance with correct classification rates of 82% and 92% in prediction sets for the diagnosis of different N application rates and N status, respectively. The PLSR model for leaf N content (LNC) showed excellent performance, with correlation coefficients of 0.924, root mean square error of 0.209, and residual predictive deviation of 2.686 in the prediction set. In addition, the important wavebands of the PLSR model were interpreted based on regression coefficients. CONCLUSION: Overall, our results suggest that the hyperspectral imaging technique can be an effective and accurate tool for qualitative and quantitative diagnosis of N status in tea plants.