In recent years, laser-induced breakdown spectroscopy (LIBS) combined with machine learning methods has become a hot research topic for detecting malignant tumors. For gliomas with infiltrative features, the tumor boundary is difficult to identify during surgery. To improve the survival time of patients, surgeons often perform extended resection, potentially damaging functional brain areas. Therefore, it is crucial to help surgeons quickly and accurately identify tumor resection boundaries during surgery. In this paper, simulation experiments are conducted using isolated tissues, proposing a polarization-resolved LIBS (PRLIBS) spectral fusion method to boost the accuracy of glioma boundary tissue detection. First, the polarization effect of the plasma emission is analyzed using the Stokes parameters, and it is found that the plasma emission belonged to partially polarized light. To better exploit the polarization information of the plasma, the polarized spectra from the four channels are fused to build a machine learning model. Comparing to classification models using LIBS intensity spectra, polarization parameters, and single-channel polarization spectra, the PRLIBS fusion model exhibits superior classification performance. The correct classification rate (CCR) of support vector machine (SVM) model is 99.05% for the training set and 89% for the test set, respectively. In the future, the PRLIBS spectra fusion method proposed in this research can be used for glioma boundary tissue identification.