Fast detection of water loss and hardness for cucumber using hyperspectral imaging technology

Hardness and water loss are the most important determining factors of the freshness of fruits and vegetables. In order to solve the defects of traditional detection methods, Hyperspectral imaging technology was investigated for fast determination of hardness and water loss of cucumber. The standard normal variate and Savitzky-Golay smoothing preprocessing methods were compared, and then optimal wavelengths were selected by competitive adaptive weighting sampling (CARS). 29 characteristic wavelengths for hardness and 42 characteristic wavelengths for water loss were selected by CARS, respectively. The partial least squares regression (PLSR) prediction models were developed based on the optimal characteristic wavelengths and the full spectrum, respectively. The results of the hardness and water loss PLSR model based on the optimal wavelengths (R-2 = 0.9420 and RMSE = 19.5088; R-2 = 0.8218 and RMSE = 1.0132) were better than those based on the full bands. Furthermore, visualized maps of hardness and water loss were built based on the generated model function, showing that the hardness and water loss change with prolonged storage time.