Estimating cadmium content in lettuce leaves based on deep brief network and hyperspectral imaging technology

Cadmium is a big threat to human health, so it is necessary to find adaptable methods to detect the cadmium content in vegetables. This study was conducted to estimate cadmium content in lettuce leaves using hyperspectral imaging system in the range of 431-961 nm covering 618 bands. Then a total of 1,260 average spectral data was calculated from region of interest (ROI). Afterwards, deep brief network (DBN) optimized by particle swarm optimization (PSO) was proposed to predict cadmium content in lettuce leaves. Besides, traditional models such as "support vector regression based on successive projections algorithm" (SPA-SVR) were also established as comparison. Finally, results shown that PSO-DBN model had the optimal performance with R-p(2) = 0.9234, RMSEP = 0.5423 mg/kg, RPDp = 3.5894 for cadmium content prediction. Therefore, it is possible to realize rapid and nondestructive detection of cadmium content in lettuce leaves by the combination of hyperspectral imaging technology and deep brief network. Practical Application Lettuce is easy to absorb heavy metals from the environment. Using hyperspectral imaging technology combined with deep brief network can realize the detection of cadmium content in lettuce leaves quickly and effectively. This work can provide an effective analysis method for study on heavy metal in vegetables.