Aiming to explore the potential of near-infrared (NIR) technology in detecting adulterated oiltea camellia seed oil, the corn oil, peanut oil, rapeseed oil and soybean oil from 12 different production areas were used as adulteration oil, and the oil-tea camellia seed oil from five different production areas were used as adulterated oil. Totally 455 adulterated oil-tea camellia seed oil samples at the adulterated mass fractions of 0, 1%, 3%, 6%, 10%, 15% and 20% were prepared. The NIR spectra of the prepared samples were obtained at the wavelength range of 833~2500nm. After the collected NIR spectra were pretreated by multiple scatter correction method, the samples were divided into a calibration set and a validation set according to the ratio of 2∶1 by using the Kennard-Stone sample partitioning method. Furthermore, successive projections algorithm (SPA), uninformative variable elimination and competitive adaptive reweighted sampling were used to extract the characteristic wavelengths (CWs) representing the adulterated oil-tea camellia seed oil samples from the investigated whole spectra. Then the support vector machine (SVM) and random forest (RF) classification models were established based on full spectra and extracted CWs. The results showed that the SVM model had higher true positive rates, while the RF model had better true negative rates. The established RF model based on the extracted nine CWs by using SPA had the highest recognition accuracy rate of 99.34%. Moreover, the recognition accuracy rate of the model was 94.74% for the adulterated oil-tea camellia seed oil samples whose adulterated mass fraction was 1%, and reached 100% for the adulterated oil samples whose adulterated mass fraction was equal to and greater than 3%. The research result provided basic data for the development of a portable detector for adulterated oil-tea camellia seed oil.