Search

Nitrogen fertilizer is necessary to improve yield and quality of lettuce. Spectroscopy is one of the most effective techniques used to detect crop nitrogen content. In this study, canopy reflectance spectra were acquired under five levels of nitrogen, and then were Savitzky-Golay smoothed, the first-order derivative spectra were calculated from the smoothed spectra to eliminate noise effects. Backward interval partial least squares (BiPLS), genetic algorithm (GA) and successive projections algorithm (SPA) were combined to select the efficient wavelengths. The number of variables was decreased from 2,151 to 8. The optimal intervals or variables were used to build multivariable linear regression (MLR) model, radial basis function neural network (RBFNN) models and extreme learning machine (ELM) models. This work proved that the results of BiPLS-GA-SPA-ELM model was superior to others with RMSEC was 0.241 6%, Rc was 0.934 6, RMSEP was 0.284 2% and Rp was 0.921 8. Our research results may provide a foundation for nutrition regulation and developing instrument.

In order to facilitate lettuce fertilization in an economically rational way, an intelligent method to identify lettuce leaf nitrogen levels was studied. Lettuce samples of different nitrogen levels were cultivated in greenhouse with soilless cultivation method. In a particular growth period, the lettuce samples in various nitrogen levels were collected, then the FieldSpech3 spectrometer was used to acquire the hyperspectral data of the cultivated lettuce leaves. As there were much noise and redundant information in original hyperspectral data, standard normal variate transformation (SNV) was used to reduce the noise of the original hyperspectral data in this paper, then the principal component waves were extracted by principal component analysis (PCA). While K nearest neighbor (KNN) and support vector machine (SVM) were used for classification studies on the processed hyperspectra data respectively, adaptive boosting (Adaboost) was introduced into the two classifiers as it could improve the classification performance of weak classifiers, then Adaboost-KNN and Adaboost-SVM, the two integrated classification algorithms, were proposed. At last, the four classification algorithms were used for classification and identification of the same test sample data respectively, with the results showing that the classification accuracies of KNN, SVM, Adaboost-KNN and Adaboost-SVM were high up to 74.68%, 87.34%, 100% and 100%, among which the classification accuracies of Adaboost-KNN and Adaboost-SVM proposed in this paper were both good, and the stability of Adaboost-SVM was the best. Therefore, Adaboost-SVM used as a modeling method is suitable for the identification of lettuce leaf nitrogen level based on hyperspectrum, and it can also be used for reference to identify the nutrient elements of other crops in nondestructive testing methods.

To monitor tea tree growth and nitrogen nutrition in tea leaves, visible-near infrared spectroscopy was used to determine total nitrogen content. One hundred eleven fresh tea leaves of different nitrogen levels were sampled according to different tea type, plant age, leaf age, leaf position and soil nutrients, which covered a wide range of nitrogen content. Visible-near infrared reflectance spectra were scanned under the sunlight with a portable spectroradiometer (ASD FieldSpec 3) in field. The software of NIRSA developed by Jiangsu University was used to establish the calibration models and prediction models, which included spectra data editing, preprocessing, sample analysis, spectrogram comparison, calibration model and prediction model, analysis reporting and system configuration Eighty six samples were used to establish the calibration model with the preprocessing of first/second-order derivative plus moving average filter and the algorithm of PLS regression, stepwise regression, principal component regression, PLS regression plus artificial neural network and so on The result shows that the PLS regression calibration model with 7 principal component factors after the preprocessing of first-order derivative plus moving average filter is the best and correspondingly the root mean square error of calibration is 0. 973. Twenty five unknown samples were used to establish the prediction model and the correlation coefficient between predicted values and real values is 0.8881, while the root mean square error of prediction is 0. 130 4 with the mean relative error of 4.339%. Therefore, visible-near infrared spectroscopy has a huge potential for the determination of total nitrogen content in fresh tea leaves in a rapid and nondestructive way. Consequently, the technique can be significant to monitoring the tea tree growth and fertilization management.