Detection of chlorophyll content distribution in cucumber leaves based on hyperspectral imaging

Abstract: The content and distribution of chlorophyll in leaves are important indicators of nutrition information in plants. The objective of this study was to investigate the spectral behavior of the relationship between reflectance and chlorophyll content and to develop a technique for non-destructive chlorophyll estimation and distribution in leaves by using hyperspectral images. The hyperspectral imaging data cube of cucumber (Cucumissativus) leaves in the range of 450-850 nm were selected and preprocessed. A rectangle mesophyll about 100×200 pixels between the second and the third branch left of the main vein was selected as the region of interest (ROI). Spectra information of characteristic bands was extracted and used to set a model with measured chlorophyll content (spectra region extracted corresponding to region chlorophyll measured). The existing modeling methods, such as artificial neural networks (ANN), support vector machines (SVM), etc., can be used to achieve better results but are inconvenient for online applications due to the introduction of sophisticated algorithms. As an operation result of multiple spectrum values (addition, subtraction, multiplication, and division, combined with linear or nonlinear ways), vegetation indices, which play a role in indicating growth and biomass of vegetation, are significant in simplifying the model. Eight representative optical indices (or signatures), which were proposed as a function of the associated reflectance (Rλ) at the special wavelength (λ) nm, were used to predict the total chlorophyll content in cucumber leaves. Finally, (R695-705)?1?(R750-800)?1was identified as an optimum index, predicting the content of chlorophyll fairly well. The correlation coefficients of each model for calibration data set (rc) and validation data set (rp) were 0.8410 and 0.8286, while RMSEC (root mean square error of calibration) and RMSEP (root mean square error of predication) were the smallest (0.2045 mg/g and 0.2190 mg/g). The optimal model showed good stability and robustness due to two major advantages, namely the effects of "red edge" and baseline removal. On one hand, two feature bands (695-705 and 750-800 nm) of the model can be used to develop a kind of portable multispectral device. On the other hand, according to the model, chlorophyll content of the leaf was estimated at every pixel. A pseudo-color map was used to describe the law of chlorophyll distribution. On the map, it is evident that the content of chlorophyll is more in the mesophyll around the veins than in the veins. The edge is seen as less than the middle of the leaf, which is consistent with the actual distribution in the leaf. That is to say, it is a feasible analysis of chlorophyll content and distribution in cucumber leaves via the technique of hyperspectral images. Our results indicated that hyperspectral imaging was considerable for predicting chlorophyll content in leaves, thus allowing the chlorophyll content to be non-destructively detected in situ in living plant samples. In addition, the distribution map can also be used to analyze the accumulation of chlorophyll in spatial plants. Besides, it is easy to facilitate monitoring distribution and variation of chlorophyll in the tissues of plants. Further studies will provide a reliable way for processes that use photosynthetic pigments to participate in such as biochemical pathway, plant growth, and mechanisms of aging.