Abstract: This study aimed to explore the feasibility of predicting the vitamin C (VC) content in “Lingwu changzao” jujubes using hyperspectral imaging and to find the best prediction model. Hyperspectral images of 100 jujube samples were collected in the wavelength range of 400 to 1 000 nm. Genetic algorithm (GA), successive projection algorithm (SPA) and competitive adaptive reweighted sampling (CARS) algorithm were used to extract the characteristic wavelengths from the original spectral data. Partial least squares (PLS) and least squares support vector machine (LSSVM) were separately used to establish VC prediction models based on the full and characteristic spectra. The results showed that standard normal variate (SNV) transformation was the best preprocessing approach. The cross validation correlation coefficient (Rcv) of the PLS model was 0.839 5, and the root mean square error of cross validation (RMSECV) was 16.248 2. GA, SPA and CARS methods were used to select 12, 5, and 26 characteristic wavelengths. The PLS model based on CARS method was the best among the models developed, and its Rc, Rp, RMSEC and RMSEP values were 0.896 2, 0.889 2, 10.746 2%, and 12.145 3%, respectively. These results confirmed the feasibility of using hyperspectral imaging for the non-destructive detection of VC content in “Lingwu Changzao” jujubes.

Key words: “Lingwu Changzao” jujubes, VC, visible near infrared hyperspectral imaging, non-destructive testing