Abstract:

In this study, traceability models for wolfberry from different geographic origins were established by using near
infrared spectroscopy combined with cluster class independent soft model method (SIMCA). Different spectral preprocessing
methods were compared for modeling. The results showed that, in the wavelength range of 950 to 1650 nm, obvious
characteristic absorption peaks at 1135, 1175, 1235, 1335, 1415, 1395 nm and 1535 nm were observed after second order
derivative processing, five-point smoothing and SNV processing of the original spectra. When the number of principal components
for each of the 8 growing areas, Xinjiang, Zhongning, Gansu, Qinghai, Nanliang, Huinong, Guyuan and Inner Mongolia
was 3, the traceability models established by the SIMCA method were the best. At the significance level of 10%, 95%,
85%, 95%, 95%, 80%, 80%, 95% and 95% of unknown samples from these growing areas were recognized and only 2.86%,
14.28%, 2.86%, 0%, 5.72%, 17.13%, 0% and 2.86% were incorrectly judged. The model for samples from Guyuan was the
best, whereas that for Huinong was the worst.

Key words: near infrared spectrum, cluster class independent soft mode method, Lycium barbarum L., origin traceability, models