This study utilized reflection terahertz (THz) spectroscopic imaging technology and petrological characteristics to investigate the pore distribution in 10 tight sandstone samples from different well locations and depths, focusing on the Fuyu oil reservoir in the Songliao Basin. The results showed significant variations in THz reflectivity due to non-uniform pore distribution. By employing a BP neural network, accurate quantification of porosity was achieved, along with the development of an evaluation model based on effective medium theory for measuring porosity levels. This approach enabled quantitative analysis of overall porosity and visualization of pore distribution within samples, providing valuable insights into reservoir fluid flow mechanisms and optimization of exploitation plan designs.