Abstract: An optical emission spectroscopy (OES) method with a non-invasive measurement capability, without inducing disturbance to the discharge, represents an effective method for material monitoring. However, when the OES method is employed to monitor the trace erosion product within the ceramic channel of a Hall thruster, it becomes challenging to distinguish between signal and noise. In this study, we propose a model filtering method based on the signal characteristics of the Hall thruster plume spectrometer. This method integrates the slit imaging and spectral resolution features of the spectrometer. Employing this method, we extract the spectral signals of the erosion product and working gas from the Hall thruster under different operating conditions. The results indicate that our new method performs comparably to the traditional method without model filtering when extracting atom signals from strong xenon working gas. However, for trace amounts of the erosion product, our approach significantly enhances the signal-to-noise ratio (SNR), enabling the identification of extremely weak spectral signals even under low mass flow rate and low-voltage conditions. We obtain boron atom concentration of 3.91×10⁻³ kg/m³ at a mass flow rate of 4×10⁻⁷ kg/s and voltage of 200 V while monitoring a wider range of thruster operating conditions. The new method proposed in this study is suitable for monitoring other low-concentration elements, making it valuable for materials processing, environmental monitoring and space propulsion applications.