Autonomy and intelligence are crucial for the development and evolution of Unmanned Aerial Vehicles (UAVs). The heterogeneous computing platform provides essential computing power for UAVs’ autonomous tasks. However, as UAV autonomy evolves, traditional heterogeneous processors with single architectural designs increasingly fall short of meeting the diverse computing demands of intelligent tasks. To address this challenge, we have introduced an innovative airborne heterogeneous fusion computing platform that seamlessly integrates the architectural benefits of multiple processors and computing accelerators. This fusion approach ensures the platform can handle a wide range of computing tasks by leveraging the unique capabilities of each accelerator type. Additionally, a task assignment model based on the Support Vector Machine (SVM) algorithm is proposed to achieve optimal allocation of tasks across the platform, enhancing efficiency and performance. A case study on real-time anomaly detection in flight data demonstrates the approach’s effectiveness. The experimental results show that the proposed method significantly enhances the efficiency of airborne computing tasks, highlighting the potential of this advanced computing platform in future UAV applications.