In the scenario of air-ground collaborative autonomous operations, the collision risk of aircraft was assessed to ensure operational safety. The collision risk of parallel routes under autonomous flight was studied based on human factors and the performance of CNS systems. Firstly, an in-depth analysis of the characteristics of autonomous operation scenarios was performed. The Cognitive Reliability and Error Analysis Method (CREAM) model was improved by incorporating cognitive behavior and interactions between pilots and air traffic controllers to assess human reliability in autonomous operations. Subsequently, combined with the CNS performance and human factors, a collision risk model for autonomous operation aircraft was established. The lateral, longitudinal, vertical, and overall collision risks of the aircraft were evaluated separately. The feasibility of this method was validated through a parallel route example, and the impact patterns of human reliability and CNS performance on the collision risk of parallel routes were analyzed. The results indicate that the collision risk values in all three directions and the overall risk remain below the target safety level (5×10^-9). Furthermore, it is demonstrated that improving human cognitive reliability and CNS performance contributes to reducing the spacing between parallel routes, thereby enhancing the safety of autonomous operations.