In fully autonomous tunnel inspection, only obstacle feature points can be used for path planning, resulting in low inspection efficiency. One of the key technologies of fully autonomous tunnel inspection UAV is obstacle avoidance path planning technology. Therefore, a fully autonomous tunnel inspection UAV obstacle avoidance path planning technology based on SLAM and IMU fusion is proposed. The node image coordinates and depth information are obtained through visual SLAM, and the coordinates are unified to the same coordinate system through coordinate transformation to construct a fully autonomous tunnel inspection node graph. Based on IMU fusion, the feature point search range is limited to the reasonable area of the current motion trend of the UAV, and a comprehensive evaluation matrix is constructed to extract obstacle feature points in the tunnel. The initially extracted obstacle feature points are preprocessed, and the key and reliable feature points for obstacle avoidance path planning are screened out by setting thresholds and Euclidean distance constraints, and the feature point cost is calculated by comprehensively considering the cumulative cost and heuristic cost. By constructing a dynamic model of the tunnel environment, using a path search algorithm fused with IMU, designing a comprehensive cost function to guide path generation, and real-time monitoring of environmental changes and triggering a path replanning mechanism, dynamic obstacle avoidance path planning of UAVs in complex tunnel environments is realized. Experimental results show that the proposed technology plans routes that are more direct and shorter, can flexibly respond to sudden obstacles, and the time cost of the proposed technology path is always lower than other methods, which fully proves that the technology can improve inspection efficiency.