关键词: 金银花, 高光谱成像技术, 硫含量, 快速检测

Abstract: For rapid and non-destructive detection of sulfur content in honeysuckle, the flowers of Lonicera japonica Thunb., hyperspectral imaging technology combined with chemometrics was applied to develop a predictive model for detecting sulfur-fumigated honeysuckle with different sulfur concentrations. Hyperspectral images of non-fumigated and sulfur-fumigated honeysuckle samples with four concentration gradients of 0%, 0.5%, 1% and 1.5% on a fresh mass basis were collected and preprocessed by Savitzky-Golay smoothing filter (S_G filter), multiple scatter correct (MSC) or standard normal variate transformation (SNV). S_G filter was selected as the optimal pretreatment method. Subsequently, the processed spectral data were used to establish models using either fisher discriminant analysis (FDA) or kernel Fisher discriminant analysis (KFDA), and the results showed that KFDA had a better discrimination accuracy of 98.2%. Considering that the full-range spectral data contain a great deal of redundancy, the characteristic wavelengths were extracted by three different methods, regression coefficients (RC), Wilks criterion and RC-Wilks. As a result, the discriminant models, RC-KFDA, Wilks-KFDA and RC-Wilks-KFDA were developed. A comparison was made between these models, and the RC-Wilks-KFDA model was found to be the best one with the highest discrimination accuracy of 100%, good classification efficiency and short running time of 0.69 s. Therefore, the S_G-RC-Wilks-KFDA model could allow fast, effective and non-destructive detection of sulfur content in honeysuckle.

Key words: honeysuckle, hyperspectral imaging, sulfur content, rapid detection