Reserch On Determination Of Heavy Metal Cadmium Content In Oilseed Rape Leaves Based On Hyperspectral Image Technology

Due to factors such as sewage discharge,pesticide fertilization,and metal smelting,heavy metal pollution of soil in China is serious,especially cadmium pollution.Oilseed rape is a common leafy vegetable that has a strong adsorption effect on cadmium in the soil,which can cause the cadmium content in the oilseed rape to exceed the standard and generate toxic compounds.This situation will not only seriously restricts the growth and development of oilseed rape,but also toxic substances can endanger the health of consumers through the food chain.Therefore,detecting the cadmium content in oilseed rape leaves is of great significance for promoting the development of oilseed rape industry and ensuring human health.The traditional detection of cadmium content in oilseed rape leaves is mainly based on laboratory testing.Although the accuracy is high,the experimental process is destructive and time-consuming.Hyperspectral image technology,as a new nondestructive testing technology,combines traditional image technology and spectral technology,that can provide rich image and spectral information.In this paper,using hyperspectral image technology and chemometrics methods,the nondestructive and high precision detection of cadmium content was studied,which provided theoretical support for the research and development of the instrument for detecting cadmium content in oilseed rape leaves.The main research contents and conclusions are as follows:(1)Collection of experimental data and analysis of detection mechanism.The hyperspectral imaging system was used to obtain the hyperspectral images of oilseed rape leaves under six different cadmium concentration gradients,selected the whole oilseed rape leaf as the region of interest,and calculated its average value as the original spectral data of leaf sample,and extracted image features(color and texture features)from hyperspectral images.The average spectral curves of oilseed rape leaves under different cadmium concentration gradients were analyzed,and the correlation between the changes of cadmium content,chlorophyll content,and internal cell structure and spectral characteristics was discussed.The feasibility of using hyperspectral image technology to detect cadmium content in oilseed rape leaves was verified.(2)Research on spectral data preprocessing and spectral and image feature extraction.The combination of SG smoothing and multiple scattering correction(SGMSC)was used to preprocess the original spectral data to reduce the impact of noise and improve the signal-to-noise ratio.Based on the advantages of iteratively retains informative variables(IRIV)and successive projections algorithm(SPA),a combination of the two algorithms(IRIV-SPA)was proposed to extract the features of spectral and image information,and compared with the single IRIV and SPA.The three methods can effectively extract the characteristic variables related to cadmium content in oilseed rape leaves.Among them,IRIV-SPA had the best information extraction effect.It can ensure that the collinearity between variables was minimal,and the number of extracted feature variables was minimal,with 10 feature wavelength variables and 5 image feature variables.(3)Establishment of cadmium content detection model in oilseed rape leaves.Support vector regression(SVR)model was used to model and analyze spectral information,image information,and spectral and image fusion information.The results showed that the SVR model based on fusion features extracted by IRIV-SPA had the best prediction effect.In order to achieve the goal of high-precision detection,particle swarm optimization(PSO)was used to optimize the SVR model,in view of the defect that PSO is easy to fall into local optimal,fireworks algorithm(FWA)was introduced to improve it.The improved firework particle swarm optimization algorithm(FWA-PSO)was used to optimize the parameters of SVR,and the optimization effect was compared with the unimproved PSO,the traditional grey wolf optimization(GWO)and ant colony optimization(ACO).The results showed that the FWA-PSO-SVR model had the highest prediction accuracy and stability,R_p~2 was 0.949,RMSEP was 0.072 mg/kg.The test results showed that FWA-PSO had a certain effect on SVR optimization,and can improve the prediction accuracy and stability of the model.Therefore,the IRIV-SPA-FWA-PSO-SVR model based on fusion features can achieve non-destructive and accurate detection of cadmium content in oilseed rape leaves.This article verified the feasibility of using hyperspectral image technology to detect cadmium content in oilseed rape leaves,and combined with chemometrics methods to achieve non-destructive and high-precision detection of cadmium content in oilseed rape leaves.It provides a reference for the detection of heavy metals in other crops,and also provides theoretical basis and support for the development of a detection instrument for cadmium content in oilseed rape leaves.