Hyperspectral characteristics and quantitative analysis of leaf chlorophyll by reflectance spectroscopy based on a genetic algorithm in combination with partial least squares regression

Abstract The precise and nondestructive detection of leaf chlorophyll content is one key to assessing the health status of crops. The objective of this study was to develop a precision method for determining the leaf chlorophyll content in rape. A genetic algorithm (GA) combined with the partial least squares (PLS) method was used to establish a chlorophyll content PLS regression estimation model based on screening the characteristic spectral regions of chlorophyll. The results show that the characteristic bands of chlorophyll in rape are 510~535, 675~695, 905~965, 1025~1225, 1165~1175, 1295~1385, 1495~1765, 1875~1895, 1970~2145, and 2179~2185 nm. Based on the characteristics of each input spectrum, the Rv2 and RPD values of the best model reached 0.97 and 5.41, respectively. This represented an increase of 0.20 and 3.42, respectively, over these values for the original full-spectrum model. The best model also achieved an RMSEP of 2.63 mg g-1, which was only 3.59% of the total sample average and was 3.78 mg g-1 less than that of the original full-spectrum model. Therefore, the best model provided good prediction accuracy for the chlorophyll content of rape. The model based on the Log (1/R) spectral transformation performed best in terms of prediction accuracy. The genetic algorithm combined with the partial least squares method (GA-PLS) can effectively screen the characteristic bands of rape chlorophyll, reduce the number of variables in the model, and produce high estimation accuracy.