Nondestructive diagnostics of nitrogen deficiency by cucumber leaf chlorophyll distribution map based on near infrared hyperspectral imaging

Abstract

Chlorophyll concentration distribution map of cucumber leaf given by near infrared hyperspectral imaging was used to determine nitrogen (N) deficiency. 120 fresh leaves picked from N deficient and control plants were used for hyperspectral image collection and chlorophyll concentration determination. Firstly, after hyperspectral image acquisition and pre-processing, the average spectra obtained from the region of interest (ROI) in leaf hyperspectral images were extracted. Secondly, principal component analysis (PCA) was performed on the average spectrum to reduce the dimension along the wavelength axis. Thirdly, multi-linear regression (MLR) was used to build calibration models relating the spectra and chlorophyll concentration which determined by high performance liquid chromatography (HPLC). The calibration models were applied to an independent prediction set for validation. Chlorophyll concentration was reasonably well predicted with a high correlation (R=0.8712). Finally, the calibration model was used to predict the chlorophyll concentration of each pixel in the hyperspectral image. Therefore, distribution maps of chlorophyll concentration on the N deficient and control cucumber leaves were obtained. Obvious differences could be directly observed from the chlorophyll distribution maps between N deficient leaves and control leaves. Depend on the maps, the mean value of chlorophyll concentration were calculated at lower nodes for N deficiency diagnostics. 12mg/g was defined as a threshold for the lower nodes leaf in diagnostics of N deficiency. Our results indicate that hyperspectral imaging exhibits considerable promise for nondestructive diagnostics of N deficiency in cucumber plant.