Retrieval of pigment contents in rice leaves and panicles using hyperspectral data by artificial neuron network models

In this paper, artificial neuron network (ANN) models were established to retrieve chlorophyll a, chlorophyll h, total chloropuhyll and carotenoids pigment contents in rice leaves and panicles using hyperspectral data. The experimental data used in this study derived from two experiments with different nitrogen levels carried out at Zhejiang University of China in 2002. The hyperspectral reflectance of the leaves and panicles at different stages was measured indoors with a ASD 2500 spectroradiometer. The total numbers of samples were 860 Cor leaves and 196 for panicles at different growth stages. The relative reflectance at green peak position(GPP), at 430 and 660nm for chlorophyll a, 460 and 640nm for chlorophyll b and 470nm for carotenoids, which were calculated by the continuum removal method, as well as the red edge position (REP),were selected as inputs of the BPN models respectively. Compared with multiple regression models, the ANN models showed significantly stronger ability in predicting the pigments contents of rice in this study.