With the rapid development of the marine transportation industry and the offshore oil exploitation industry, oil spill pollution is becoming increasingly serious, posing a great threat to the marine environment and ecological balance. Therefore, the treatment and improvement of oil spill pollution have become an urgent and important work in marine environmental protection engineering. The identification of oil spills in different states is the basis and key to solving the problem of oil spill pollution. The oil spill on the sea mainly includes two different stages: non-emulsification and emulsification. The former is in the form of oil film with different thicknesses, while the latter is in oil spill emulsion with different oil-water ratios. The oil spill in different states has different element compositions: the oil film is a pure oil molecule, the emulsified oil spill is an oil-water mixed structure, and the fluorescent group is formed. Under the action of the laser, it has its own characteristic fluorescence spectrum information, and different states show obvious fluorescence spectrum differences. The shape feature of the spectral curve is an external manifestation of fluorescent substances’ physical and chemical properties, so analyzing and comparing certain spectral parameters from the shape feature of the spectrum can achieve the purpose and effect of substance classification and species identification. In order to realize the rapid classification and identification of different states of oil spills on the sea, the LIF detection system was built to collect the fluorescence spectra of common oil products in different states. The comparison of the spectral curves shows that the spectrum in the emulsification stage will show a series of characteristics, such as the increase in the number of fluorescence peaks, the change of fluorescence intensity, the shift of fluorescence peak position and so on. According to the principle of apparent statistics, the mean value, standard deviation, kurtosis coefficient, spectral linewidth, curve slope and other characteristic parameters of the spectrum are extracted, and these characteristic values are used for cluster analysis. The results show that the cluster analysis results of oil spills based on laser-induced fluorescence spectrum are consistent with the actual oil spill status. Based on the premise of known oil species, the classification method can better identify different oil spill states on the sea. Therefore, this method can provide a new idea for identifying oil spills on the sea and lay a foundation for the improvement of LIF technology detection quality and application level.