In order to ensure the quality of seedlings and provide healthy and robust seedlings to meet the needs of large-scale and standardization of modern rapeseed industry, a 21d temperature stress experiment of rapeseed seedling was carried out. The aim was to study the identification of robust seedlings of rape under temperature stress using hyperspectral imaging technology. Firstly, the sensitive bands of temperature stress were extracted by spectral reflectance and continuous wavelet transform. And then the continuous projection algorithm and continuous wavelet transform-stepwise discriminant analysis were respectively used to extract characteristic wavelengths from sensitive bands of temperature stress. The waveband features and spectral features of rapeseed seedlings were analyzed with time. A total of seven features were selected, including the curve area at band MA and tangent eigenvalue tanθ of 554~714nm, the reflectance value at 1213nm and 1567nm, wavelet feature w(9, 967), w(13, 1213) and w(7, 1567) to establish a multi-feature fusion Fisher discriminant model. The results showed that the average classification accuracy of the model was 88.68%, and the best detection accuracy reached 95.56% at the three-leaf stage, which could better distinguish the temperature stressed rape seedlings and provide a reference for the rapid detection of robust rape seedlings based on hyperspectral technology.