Hyperspectral model based on genetic algorithm and SA-1DCNN for predicting Chinese cabbage chlorophyll content

Chinese cabbage (Brassica pekinensis L.) is a leafy green vegetable, which is widely grown and consumed worldwide. Rapidly determining chlorophyll content is crucial for the effective and efficient breeding and cultivation of Chinese cabbage. Hyperspectral techniques have been widely used for studies concerning chlorophyll content; however, few studies have been conducted for different periods and species of Chinese cabbage leaves. In the present study, we used a Chinese cabbage translocation line material with rich variation. Samples were acquired for covering all periods of Chinese cabbage nutritional growth. In this study, a hyperspectral data preprocessing method was combined with a genetic algorithm for feature wavelength extraction. A self-adjusted prediction model based on a One-Dimensional Convolutional Neural Network (1D-CNN) was designed to predict four chlorophyll contents: Chlorophyll a (Chla), Chlorophyll b (Chlb), total Chlorophyll (Chl), and SPAD. Compared to the traditional machine learning model, partial least squares regression (PLSR), the self-adjusted model exhibited optimal performance. In the test set, the R2 of the SA-1DCNN model was 0.65, 0.64, 0.71 and 0.34 for each Chlorophyll (included Chla, Chlb, Chl and SPAD), respectively. The experimental results showed that hyperspectral preprocessing methods combined with a genetic algorithm (GA) and Self-adjusted One-Dimensional Convolutional Neural Network (SA-1DCNN) were effective for chlorophyll content prediction. This study provides new theoretical framework for determining the nutritional status of Chinese cabbage.