| Heavy metal cadmium pollution severely restricts the development of the vegetable industry and threatens animal and human health through the food chain.At present,the detection of heavy metal cadmium mainly depends on the laboratory chemical detection method,which are destructive to samples and complex for sample pretreatment.However,spectral detection has the disadvantages of anisotropy and one-sided information.Therefore,it is very urgent to explore and establish a rapid,non-destructive and accurate method for detecting the cadmium content in vegetables.In this paper,Lettuce as the research object,the effect of cadmium stress on lettuce leaf surface was observed by scanning electron microscopy(SEM),then the impact of cadmium stress on lettuce leaf mesophyll cells was observed by transmission electron microscope(TEM).Next,the content of heavy metal cadmium in lettuce leaves were determined by using the national standard method for determination of heavy metal cadmium(GB 5009.15-2014).Finally,the content of heavy metal cadmium in lettuce leaves was detected by using hyperspectral image technology combined with chemometrics methods.The main findings are as follows:(1)SEM and TEM were used to observe the effects of cadmium stress with different gradients on leaf tissue of lettuce.It was found that the higher cadmium stress gradient,the more obvious the stomatal closure on the surface of lettuce leaf photographed under the SEM.After cadmium stress,there will be more osmiophilic particles,destruction of chloroplast membrane system,disappearance of thylakoids,and disorder of stacking.SEM and TEM analysis of lettuce leaves subjected to cadmium stress can be used as a feasible basis for the detection of cadmium in lettuce leaves by hyperspectral imaging.(2)Spectral data and image features were extracted from hyperspectral images.ENVI software was used to select the region of lettuce leaf as the region of interest(ROI),and the average value of the spectral data in the ROI was calculated as the spectral data of the sample.MATLAB software was used to extract gray-gradient co-occurrence matrix and mean,variance,skewness,Kurtosis of R,G,B,H,S,V color feature values of lettuce leaf images.A total of 24 color feature values and 15 texture features values were obtained.(3)SG smoothing and multi-scatter correction(MSC)were used to preprocess the original spectral data.Feature bands of the pre-processed spectral data were selected by using VCPA,MASS and BOSS.The bands selected by MASS and BOSS have good modeling effects but the number of selected bands is large.Due to VCPA can select the fewer bands,VCPA was used to perform secondary feature extraction on MASS and BOSS.By comprehensively comparing the number of feature bands and the performance of the quadratic feature extraction model,it was found that the effect of the model built on the feature bands selected by the combination of VCPA and MASS is optimal,and the number of selected feature bands is 13.(4)Quantitative detection models of heavy metal cadmium in lettuce leaves of SVR,GWO-SVR and DE-GWO-SVR was established based on feature spectral data and fusion information.The results show that DE-GWO-SVR model based on fusion information has the best performance.The coefficient of determination of training set is 0.9723,the root mean square error of training set is 0.0096 mg/kg,and the correlation coefficient of forecasting set is 0.9512.The root mean square error is 0.0234 mg/kg.(5)The feature bands of each pixel of whole ROI image extracted by MASS-VCPA was plugged into DE-GWO-SVR forecasting model.Then the cadmium content of each pixel value was predicted.Finally,heavy metal cadmium content in lettuce leaves was realized.In this paper,the quantitative analysis of cadmium content in lettuce leaves was carried out by hyperspectral image technology and chemometrics,and was realized the visualization of the cadmium content in the lettuce leaves.This research can provide technical support to ensure high yield and high quality of lettuce. |
PDF Full Text Request