In order to seek a method which can detect delinted cottonseeds vigor rapidly and non-destructively. Hyperspectral imaging is an emerging technique that is applied in detection of agricultural products in recent years. Experiments of three varieties of delinted cottonseeds with different aging degree were conducted, including Xin Luzao 50, Xin Luzao 57, Xin Luzao 62. The hyperspectral image of 810 grain of delinted cottonseeds in the range of 450~1 000 nm was collected with hypersectral imaging system. Different pretreatment methods were combined and chauvenet detection method excluding outlier was applied to establish a partial least squares (PLS) , Stepwise Multiple Linear Regression (SMLR), Principal Component Regression (PCR) model. The results shows that the best combination of two kinds of pretreatment methods were standard normal variate (SNV), Savitzky-Golay (S-G) smoothing, First derivative and standard normal variate (SNV), First derivative, Norris Smoothing. After preprocessing, combined with the range of 480~530, 650~980 nm to establish three different kinds of models. The partial least squares (PLS) model effect is the best. As for the prediction set and calibration set of PLS model of conductivity, Xin Luzao 50, Xin Luzao 57, Xin Luzao 62, correlation coefficient of prediction set and correlation coefficient of calibration set were 0.92, 0.95, 0.92, 0.90, 0.89, 0.90. Xin Luzao 50, Xin Luzao 57, Xin Luzao 62 root mean squared error of prediction (RMSEP) and root mean squared error of calibration (RMSEC) were 44.3, 38.4, 37.8, 46.5, 43.5 and 40.8 μS·cm-1, respectively. This paper studied the applicaiton of hyperspectral image technology to detect the vigor of delinted cottonseeds, not only provides a new method for detecting delinted cottonseed vigor, but also lays a theoretical foundation for other seed vigor test.