In the Ro-Ro ship loading decision-making process, issues such as overly simplified deck boundaries that do not align with reality, failure to comprehensively consider simultaneous loading of multiple deck levels and various vehicle types, and low deck utilization rates have been identified. To address these challenges, a multi-deck automatic loading decision-making model for Ro-Ro ships was developed, tailored to the unique characteristics of Ro-Ro ship loading. First, by integrating the characteristics of deck boundaries and obstacles, the lowest waterline method was combined with improved strategies such as deck boundary localization algorithms and obstacle detection algorithms to address issues related to irregular deck boundaries and the presence of internal obstacles. Second, a gene fragment method was proposed to group vehicles of the same type together for processing, ensuring that similar vehicles were loaded adjacent to each other, thereby relatively increasing deck utilization by 9.24%. Finally, a two-layer integer coding strategy was employed to simultaneously process multi-layer decks and vehicles, meeting the requirements for automatic loading of multi-layer decks. Through simulation verification, the proposed multi-layer deck automatic loading decision-making model achieves a deck utilization rate of approximately 90%, demonstrating good performance and effectively addressing the practical loading challenges of Ro-Ro ships.