The invention discloses a method for rapidly identifying red jujube varieties. The method comprises the following steps: respectively selecting hyperspectral data in upper, lower, left, right and middle areas of interest from each red jujube, and solving the average spectrum of the five areas of interest to serve as spectral data of the sample; measuring the maximum size of the length, width and height of the red jujubes through an MYD64A1 measurement light curtain so as to obtain an approximate volume of the red jujubes; selecting an optimal two-dimensional wavelet to perform multiscale decomposition on two-dimensional correlation spectroscopy and respectively reconstructing each layer; performing modeling prediction on each layer of reconstructed correlation spectrum and obtaining a root-mean-square error of cross validation; performing sub-model fusion through the calculated weight; and evaluating the result and performance of the multiscale-two-dimensional correlation spectrum model by predicting the root-mean-square error and correlation coefficient. According to the method disclosed by the invention, the maximum size of the length, width and height of the red jujubes is measured based on the light curtain method, and the red jujube varieties are rapidly identified by utilizing the volume difference and spectral data.