Abstract: In this paper we aim to provide a rapid, simple and accurate method for determining camellia seed oil content, by the DA7200 near infrared apparatus (made in Sweden). One hundred and fifty seeds that were collected from different origins in Guangdong province were tested. Based on the analysis data of the absorption spectra and the chemical characteristics, two models for kernel oil content determination were established, and then the accuracy of the models was evaluated. The result shows that the model with the whole seed kernel can reached to 0.88 of the correlation coefficient between the prediction and the chemically measured values, and the standard deviation of prediction is in the range of 0.91. Thus this model with the whole seed kernel can be used to measure low quantity and precious samples with the relative low accuracy. The model with the crushed seed kernel can reached to 0.98 of the correlation coefficient between the prediction and the chemically measured values, and the standard deviation of prediction is in the range of 0.33. This model is very well and can directly replace conventional method. The test results show that prediction models have high precision and good stability. The forecast results indicated that near-infrared spectroscopy for rapidly testing the oil content of dry kernel was useful and would have a very good application prospects.

Key words: near infrared spectrum, Camellia oleifera, oil content, spectral calibration model