Spectral classification of lettuce cadmium stress based on information fusion and VISSA‐GOA‐SVM algorithm

Abstract In order to study the qualitative Vis–NIR spectra detection of cadmium stress in lettuce leaves, this article uses data‐level information fusion method coupled with VISSA‐GOA‐SVM algorithm for research and analysis. There were 140 pieces of lettuce leaves under four gradients of cadmium stress, as well as a total of 560 samples of lettuce were used to collect Vis–NIR spectral information. Wavelet transform (WT), first derivative (1st Der) and second derivative (2nd Der) were used to process original Vis–NIR spectral data, respectively. Besides, the data after different pretreatments will be fused in the data‐decision level and used as a new input layer. Principal component analysis (PCA), iteratively retaining informative variables (IRIV), and variable iterative space shrinkage approach (VISSA) were used to reduce the dimensionality of the data layer, respectively. Besides, support vector machine (SVM) was used to establish a classification model. The results showed that the RTD (initial fusion of three data input layers) combined with VISSA‐GOA‐SVM (VISSA combined with grasshopper optimization support vector machine) classification model was the best, and the accuracy of the calibration and prediction were 100% and 98.57%, respectively. Application of Vis–NIR hyperspectral imaging technique to detect cadmium stress gradient in lettuce leaves provided a new method for identifying different heavy metal residues in lettuce. Practical applications It is of great significance to detect different cadmium stress levels through nondestructive testing. In order to effectively implement the rapid and nondestructive testing of lettuce leaves under different cadmium stress levels, Vis–NIR hyperspectral imaging technology coupled with data‐level fusion method was used in this article. In addition, VISSA‐GOA‐SVM model was established to detect cadmium stress level. It confirms that the Vis–NIR hyperspectral imaging technology is a feasible and effective method for discriminating different cadmium stress levels.