Abstract: To overcome the difficulty in the artificial identification of harmful algae,the advantage of the autofluorescent nature of harmful algae is employed in an automatic recognition system presented in the paper,based on the technology of double-spectrum.In the system,the pre-processing,segmentation,feature extraction,feature selection,establishment of harmful algae information database and classification can be fully conducted automatically.Besides,the back-propagation neural network is used in the recognition.Experiments show that the classification accuracy of the system is up to 90%, as compared with the human identification.