The three-dimensional fluorescence spectroscopy was combined with the self-weighted alternating three linear decomposition (SWATLD) algorithm to detect the three kinds of mixed pesticides. In acetonitrile solvent, preparation of carbaryl, metolcarb and Triazophos, three mixed solution of different concentration ratio as a measurement sample (carbaryl, metolcarb and Triazophos by three optimal excitation wavelength / emission wavelengths were 285/325, 305/345, 265/305 nm), three-dimensional fluorescence spectra were obtained by fluorescence spectrometer, and after blank subtraction, excitation and emission correction, the influence of instrument error was effectively removed and real scattering spectra of samples were obtained. In this paper, the self-weight alternating three linear decomposition algorithm is applied to analyze the measured spectral data, and the average recovery rate of the three pesticides is 96.9%±1.9%, 99.8%±1.0% and 100.8%±3.2%. According to the results of the SWATLD algorithm, calculation of three kinds of pesticide of the root mean square error (RMSEP) value is 0.616×10-2, 0.539×10-2 and 0.374×10-2 μg·mL-1, lower than the parallel factor analysis (PARAFAC) method to predict the results of the RMSEP value, and the minimum detection limit was 0.005~0.022 μg·mL-1 range. Compared with the PARAFAC algorithm, the advantages of the SWATLD algorithm are highlighted. It shows that the algorithm has good decomposition ability for the three kinds of pesticide mixtures with severe spectral overlap.