In order to efficiently monitor the pesticide residues in vegetables, a detection method of pesticide residue content of imidacloprid in cabbage on fluorescence spectroscopy was proposed. Firstly, 400nm was determined of as the optimal excitation wavelength of imidacloprid by three-dimensional fluorescence spectroscopy. Afterwards, six pre-processing algorithms and two dimensionality reduction algorithms were analyzed. Multiple scattering calibration (MSC) and uninformative variable elimination (UVE) were selected as the best pre-processing and wavelength selection methods, respectively. Finally, the broad learning system (BLS) was used for fluorescence spectroscopy modeling and compared with classical models such as partial least squares regression (PLSR), support vector machine (SVM), and deep extreme learning machines (DELM). The results showed that the BLS model obtained the best prediction of imidacloprid content. The test set coefficient of determination (R2p) reached 0.949 and the root mean square error (RMSE) reached 0.347mg/kg. The research result showed that fluorescence spectroscopy combined with BLS was feasible to identify pesticide residue content, and it can provide a theoretical basis for the development of online detection system for pesticide residue content.