Aiming at the mental health problems of poor college students, this study proposes an expression recognition method based on the optimization of PCA algorithm and applies it to the mental health test framework. By comparing the differences between poor college students and norms, and using multiple linear regression and typical correlation analysis to explore the impact of educational management paths. The study also addressed the effect of different types of colleges and universities on the level of management of academic support for poor students, and found that 985 colleges and universities significantly outperformed the norm in this area, i.e. P > |t| = 0.031. This provides valuable insights into improving educational and psychological support for poor college students.