Construction of Superoxide Dismutase Activity Model of Cabbage Leaves Based on Deep Learning
Superoxide dismutase(SOD)is a key index to judge the degree of crop stress,which can be used to measure plant growth status,and also has important significance in the study of plant stress.In order to achieve rapid nondestructive detection of SOD activity in cabbage leaves,a method of hyperspectral imaging(HSI)combined with deep learning was proposed to detect cabbage leaves.In the experiment,spec-tral information of 200 cabbage leaves at different growth and development stages was collected,and the original spectra were optimized by 7 pretreatment methods through sample set division.Finally,the Gaussian Filter(GF)method was selected as the pretreatment method for SOD activity.Successive projection algorithm(SPA),uninformative variable elimination algorithm(UVE),genetic algorithm-partial least squares al-gorithm(GAPLS),competitive adaptive reweighted sampling(CARS)and interval variable iterative space shrinking analysis(IVISSA)were used to extract feature wavelengths and partial least squares regression(PLSR)model was established.PLSR,principal component regression(PCR),multiple linear regression(MLR),least square SVM(LSSVM)and convolutional neural network(CNN)models were established based on the preferred characteristic wavelength.The results showed that the 17 optimal wavelengths extracted by CARS algorithm had the best effect,and the correlation coefficients Rc and Rp values of the optimal prediction model CNN were 0.909 8 and 0.823 5,respectively.And the root-mean-square error RMSEC and RMSEP were 2.038 2 and 3.649 2,respectively.This study provided technical support for non-destructive on-line monitoring of plant growth under salt stress in the future,and had a good development prospect.
万方数据
维普
