For quick, nondestructive, and cheap testing of the volatile oil of Zanthoxylum bungeanum, Chinese prickly ash samples were selected as the experimental object and collected from Hanyuan County for hyperspectral analysis in the 400-1000-nm wavelength. Standard normal variable transformation (SNVT)was used to preprocess the spectral data and the method of iteratively retains informative variables (IRIV) was used to extract the feature variables. The regression model of extreme learning machine (ELM) was established. The following results were obtained using the model: the coefficient of determination ( RC2) and root-mean-square error of the calibration set (RMSEC) were 0.8522 and 0.3475 and the coefficient of determination ( RP2) and root-mean-square error of the prediction set (RMSEP) were 0.8365 and 0.5737. To improve the prediction performance of the model, the fruit fly optimization (FOA) algorithm was used to optimize the input weights of ELM. Finally, RC2 and RMSEC of the optimized model (IRIV-FOA-ELM) were 0.8792 and 0.3323, respectively, and RP2 and RMSEP were 0.8659 and 0.3621, respectively. The results show that the hyperspectral imaging technique can be used for the rapid nondestructive testing of the volatile oil of Z. bungeanum, providing a new method and concept for the testing of volatile oil of other agricultural products.