The invention discloses a method for detecting apple sour through the hyperspectral technology. The method comprises the following steps that hyperspectral data of apple samples is acquired through a hyperspectral sorter; corresponding taste data of the apple samples is acquired through an electronic tongue; middle 480 wavebands are selected as original spectroscopic data, after the original spectroscopic data is preprocessed, partial least squares regression modeling analysis is carried out, correlation coefficients R and root-mean-square errors (RMSE) of a calibration set and a prediction set are obtained, waveband values corresponding to the maximum value and the minimum valve of 18 local correlation coefficients are selected as feature wavebands for establishing an apple sour prediction model. According to the research method for detecting apple sour indexes through the hyperspectral technology, the feature waveband partial least squares regression model is established through the hyperspectral data and the data measured by the electronic tongue, a template is provided for following detection, and the method has the advantages of being scientific and reasonable, easy to operate, high in taste accuracy rate and the like.