This study focuses on the plasma spectroscopy phenomena during the Laser Shock Peening (LSP) process, aiming to explore the correlation between spectral characteristics and surface integrity. It proposes an online monitoring method based on plasma spectroscopy to enable intelligent monitoring of the LSP process. First, the parameters of the plasma spectroscopy acquisition system are introduced, followed by a detailed analysis of the plasma characteristics during LSP. Using the Boltzmann two-line method with N Ⅱ 500.515 nm and N Ⅱ 399.5 nm spectral lines, the apparent plasma temperature is estimated to range between 16 000 and 22 000 K. Additionally, the electron density, calculated using the Hα 656.27 nm spectral line, is approximatelyin the range of 2.287 to 3.612×10¹⁶ cm^-3. As the laser power density increases, both plasma temperature and electron density exhibit an overall increasing trend. However, fluctuations are observed due to the inherent spatiotemporal instability of the LSP process.Subsequently, the study investigates the strengthening effects of LSP on the surface integrity of γ-TiAl alloy. Experimental findings show that as the laser power density increases, the surface residual stress rises from an initial value of -61.498 to -444.224 MPa, while the surface Vickers hardness improves from 317.8 to 385.5 HV. Based on these experimental data, polynomial fitting models are developed to predict surface residual stress and Vickers hardness, using the intensity ratio of the Hα line and the N Ⅱ 500. 515 nm line as independent variables. Both models achieve determination coefficients (R²) exceeding 90%, providing a reliable foundation for quantitative predictions of surface integrity. To enable efficient online monitoring, the study proposes an end-to-end monitoring method based on a CNN-Transformer deep learning architecture. By processing plasma spectroscopy data, the model performs monitoring of the LSP process. Experimental results demonstrate a classification accuracy of 99.3%, highlighting the approach's efficiency and reliability for online monitoring. In conclusion, by integrating physical modeling and deep learning techniques, this study establishes the correlation between plasma spectroscopy and surface integrity during the LSP process, providing an innovative and reliable approach for intelligent online monitoring of LSP.