This study addresses the issue of casualty evacuation during earthquake disasters by constructing a casualty transfer model that integrates connected graphs and linear programming. The model first establishes the topological relationship of the road network using the Floyd algorithm, determining the shortest transfer routes from disaster areas to hospitals. Subsequently, through a linear programming approach with constraint settings, the optimal number of casualties to be transferred is determined to ensure the shortest transfer routes and minimal total time consumption. A genetic algorithm model is also developed for comparative validation. Taking a specific area in Hohhot as the research subject, this study simulates the transfer routes and numbers of severely injured casualties under the assumption of a strong earthquake occurring in the evening or at night, estimating the number of casualties and road traffic conditions through empirical formulas. The results demonstrate that the model can effectively solve for the minimum total time and shortest total routes required for casualty transfer, verifying its practical feasibility. Simulation experiments further confirm that the model significantly improves casualty evacuation efficiency, reduces rescue costs, and shortens rescue time. Additionally, the practical applicability and potential for future expansion of the model are discussed, aiming to provide a theoretical basis and practical reference for earthquake disaster emergency rescue.