Abstract: The photometric characteristics of satellites play an important role in analyzing their physical properties. An accurate photometric model can effectively support observation planning and strategy formulation, while also facilitating the analysis and interpretation of satellite attitude and structural characteristic changes. However, due to the rapid and complex variations in attitude and observation conditions during Low Earth Orbit (LEO) satellite transits, existing photometric models fail to accurately estimate the brightness of these satellites. To address this issue, we propose a modified spherical photometric model based on equivalent cross-sectional area, specifically designed for LEO satellites with stable attitude and orbit, and with observational data that sufficiently covers various viewing conditions. This method equates the satellite's cross-sectional area to that of a sphere with equal brightness. By incorporating the relative positional relationships among the satellite, the Sun, the observation station, and the orbital plane as key parameters, it employs machine learning algorithms to effectively fit and estimate the equivalent cross-sectional area, thereby enabling accurate photometric predictions during LEO satellite transits. To validate the effectiveness of the proposed method, we conducted performance evaluations using V-band photometric data from 7 LEO satellites observed by the 50 cm telescope at Xinglong Observatory, with comparisons made against the traditional spherical model. The results show that when sufficient observational data is available, the proposed method achieves higher accuracy and greater stability in photometric prediction, demonstrating strong reliability and broad applicability.