Bacterial biofilms widely exist on the surfaces of food processing machinery， medical equipment and the environment， and bring a huge threat to human health due to strong drug resistance. Escherichia coli，Staphylococcus aureus， and Salmonella typhimurium were used to study the feasibility of species identification and evaluate film-forming ability of different bacterial biofilms by hyperspectral fluorescence imaging technology to solve the problems of time-consuming， laborious and inefficient detection of existing biofilms. The hyperspectral fluorescence images of bacterial biofilm samples were collected， and support vector classification machine （SVC） and partial least squares discriminant analysis （PLS-DA） models based on the spectral data preprocessed by five methods were established to classify bacterial biofilms. Characteristic wavelengths were extracted by employing successive projections algorithm （SPA） and competitive adaptive reweighted sampling （CARS） respectively， and the corresponding simplified models were established. The results showed that SVC outperformed the full-wavelength and characteristic-wavelength models of identifying bacterial biofilm species than the PLS-DA， with the optimal model being None-SPA-SVC， where the classification accuracy of calibration set （CCRC） and prediction set （CCRP） were both 96.67%. In the classification and discrimination of film-forming ability of bacterial biofilm， the full-wavelength SVC models generally outperformed PLS-DA with higher classification accuracy. For the simplified models， the optimal model was SPA-SVC， with CCRC and CCRP of 100.00% and 96.67%， respectively. It is indicated that hyperspectral fluorescence imaging technology can effectively， quickly and accurately classify the types of bacterial biofilms and the film-forming ability of biofilms.