Abstract: On-line detection of insulator contamination is essential for realizing the intelligence of power equipment in the ubiquitous power internet of things strategy, and is important for the safe and reliable operation of the external insulation of power systems. Research on online, fast and accurate methods for detecting pollution is of great significance for monitoring the operating state of transmission lines. In this paper, a quantitative method for the rapid detection of laserinduced breakdown spectroscopy（LIBS） and the competitive adaptive reweighting sampling-the partial least squares regression（CARS-PLSR） of insulators was studied. The spectral data was preprocessed by different spectral preprocessing methods. The CARS was used to optimize the selected spectral variables. Then PLSR was used to establish a quantitative analysis model of the equivalent salt density and the spectral intensity of the elements tested. The results showed that the normalized preprocessing method combined with the CARS can improve the analysis accuracy of the PLSR method.