Excitation-emission matrix spectroscopy (EEMs) was widely used in food science, analytical chemistry, biochemistry and environmental science because of its wealth of information and high sensitivity. However, the Rayleigh and Raman scatterings bring difficulties to display data and quantitative analysis of EEMs. Therefore, in the early stage of data processing, eliminating the scatterings interference is significant for the popularization and application of EEMs. In previous studies, few researchers aimed to deal with the low-concentration solution and the scattering peaks overlap with the material fluorescence signals. In order to solve this problem, the paper proposed a self-adapting method to remove the scatterings of EEMs. First, the method corrects the Raman scatterings and background interference by subtracting the spectral baseline of the solvent obtained in each experiment. Then, according to the intensity and overlap degree of fluorescence signals and scattering peaks. EEMs are divided into five categories corresponding to three overlap levels: no overlap, weak overlap and serious overlap. The Rayleigh scatterings are corrected by setting to zero, piecewise cubic Hermite interpolation polynomial algorithm and Gaussian-piecewise cubic Hermite interpolation polynomial coupling algorithm, respectively. The method is based on one-dimensional interpolation, and only emission spectrums are studied, reducing the algorithms complexity and computation time. The methods effectiveness was demonstrated by the experiments on four typical organic compounds: Tyrosine, Fulvic acid, Naphthalene acetic acid and Rhodamine B. Moreover, compared with the Delaunay interpolation method, which is most used in current research. The concentration regressions are performed with the EEMs after the scatterings correction; the average of -R-squared obtained by the method is 0.9962, which is 5.04% higher than the Delaunay interpolation method. At the same time, by comparing the EEMs of Fulvic acid corrected by the method and Delaunay interpolation method, it is proved that this method can better maintain the structural characteristics of the fluorescence spectral regions and effectively avoid the occurrence of overfitting. Finally, this method is used to monitor the simulated sudden pollution water samples, which verifies that the method has good universality and practical application value. It provides a novel idea for removing scatterings interference of EEMs.