A novel thickness measurement NIR spectrometry for surface insulation coating of silicon steel based on discrete binary particle swarm optimization (DBPSO) algorithm is presented. First, we used NIR spectrometer to collect the NIR spectra of insulation coating of silicon steel, and then, DBPSO algorithm was used to select the optimal wavelength variates and composed a new spectra set. Last, the authors created the thickness quantitative analysis model using kernel partial least square algorithm. The experimental results show that the absolute error range analyzed by created model was from -0.12 to 0.19 μm, and the maximal relative error was 14.31%, which completely met the practical measurement need. The research indicates that DBPSO is effective wavelength selection methods, which can efficiently select the wavelength variates carrying more useful information, improve the analysis accuracy and speed. And the NIR spectroscopy is an effective measurement method for thickness analysis of silicon steel insulation coating.