Visualization research of moisture content in leaf lettuce leaves based on WT‐PLSR and hyperspectral imaging technology

Abstract Fast and effective visualization research of leaf lettuce leaves was particularly important in modern fine agriculture irrigation. Therefore, hyperspectral imaging technology was used to test the moisture content in leaf lettuce sample. A method involving wavelet transform coupled with partial least squares regression (WT‐PLSR) was proposed to extract characteristic wavelengths, build models, and evaluate characteristic wavelengths. Hyperspectral imaging data of 200 leaf lettuce leaves of five moisture gradients were obtained using hyperspectral imaging instrument. The whole region of leaf lettuce sample was selected as region of interest (ROI) to extract the hyperspectral data using sheffield index, image segment, and masking. To reduce noise interference, Savitzky–Golay (SG) algorithm was used to manage spectral data. Besides, WT‐PLSR algorithm was used to select effective feature wavelengths, build detection model, and evaluate characteristic wavelengths. The best prediction model of moisture content in leaf lettuce sample was PLSR model ( RMSEP = 0.1688, =0.8307) building by characteristic wavelength extracted from fourth layers of wavelet decomposition, and it was applied to achieve the distribution of moisture content in leaf lettuce leaves. Hyperspectral imaging technology coupled with WT‐PLSR algorithm can effectively realize the quantitative detection of moisture content in leaf lettuce sample, and visualizing distribution map of moisture content in leaf lettuce leaves offered a more intuitive and comprehensive assessment of moisture contents. Practical applications Well understanding moisture content in leaf lettuce leaves is very important for revelation of novel biological function and calcium (Ca) content. To facilitate more intuitively and comprehensively detect the moisture content in leaf lettuce leaves, a method involving wavelet transform coupled with partial least squares regression (WT‐PLSR) was proposed to extract characteristic wavelengths, build models, and evaluate characteristic wavelengths. It confirms that the WT‐PLSR algorithm is feasible and effective for building models of different moisture contents in leaf lettuce leaves, and visualizing distribution map of leaf lettuce leaves offered a more intuitive and comprehensive assessment of moisture contents.