Abstract: With the expansion and increased complexity of power grids, traditional power inspection faces challenges of low efficiency, high risk, and high cost. The contradiction between diversified power inspection scenarios and the performance limitations of single-type UAVs, as well as difficulties in multi-source heterogeneous data processing and multi-UAV collaborative scheduling, have constrained the effectiveness of UAVs in power inspection applications. This paper investigates real-time dynamic scheduling control technologies for precise take off and landing, develops a precision inspection scheduling platform based on multi-source heterogeneous UAVs using fusion positioning methods combining RTK with AprilTag and multi-agent reinforcement learning algorithms. The platform constructs scene feature vector models, information entropy-based data fusion frameworks, and collaborative scheduling models based on capability-task matching degrees, achieving precise positioning, dynamic route replanning, and intelligent task allocation for multiple UAVs. Practical application results show that the platform improves inspection efficiency by 280%, reduces costs by 58.3%, achieves landing precision of ±5cm in normal weather conditions while maintaining ±15cm in adverse environments, with an investment recovery period of approximately 1.8 years, providing an effective solution for intelligent power inspection.