Abstract: Gastrodia elata is one of the important medicinal herbs in Yunnan Province, and gastrodin and 4-Hydroxybenzyl Alcohol are the important active ingredients in Gastrodia elata, which are important indexes for evaluating the quality of Gastrodia elata. The detection methods of gastrodin and 4-Hydroxybenzyl Alcohol in Gastrodia elata are high performance liquid chromatography, gas chromatography, etc., which are time-consuming and complicated to operate. Near-infrared spectroscopy has the advantages of fast detection, convenience and non-pollution, and has a ?wide range of applications in quality control, quantitative as well as qualitative analysis in the field of Chinese herbal medicines. In order to realize the rapid and non-destructive detection of gastrodin and 4-Hydroxybenzyl Alcohol in Gastrodia elata, this study took Gastrodia elata from Zhaotong, Yunnan province as the experimental object, and collected spectral data in the range of 900-1700nm. Firstly, convolutional smoothing (SG) and standard normal variable transformation (SNV) are used for spectral data preprocessing, and secondly, feature wavelength extraction is carried out by CARS and IRIV, and according to the results of establishing LSSVR model based on feature wavelength, the best feature wavelength extraction method is selected. In order to improve the accuracy of the model, this study introduces an artificial rabbit intelligence algorithm to optimise the regularisation parameter γ and the kernel function density σ2 in LSSVR. And in order to evaluate the superiority of ARO, it was compared with PSO and GWO. The overall results show that ARO algorithm is superior to PSO and GWO in search speed and search ability. The best prediction models for both gastrodin and 4-Hydroxybenzyl Alcohol were CARS-ARO-LSSVR, with R2p of 0.9804 and 0.9797, and RMSEP of 0.0110 and 0.0140, respectively. Therefore, this study shows that near-infrared spectroscopy can be used for quantitative detection of active components in Gastrodia elata, which provides part of the theoretical basis for the development of rapid detection devices for Gastrodia elata.

Key words: Near infrared spectroscopy, Gastrodia elata, Least-Squares Support Vector Regression, Artificial rabbit intelligence algorithm