Long distance dike inspection requires full section coverage, high-frequency deployment, and reliable defect detection, while manual inspection and stand-alone operation are still limited by efficiency, continuity, and data consistency. Under the SL/T860-2026 framework, a UAV swarm inspection framework integrating multi-source perception, collaborative path planning, multi-source processing, improved detection and specification compliance evaluation is constructed. The system uses six VTOL fixed-wing Uavs, three multi-rotor Uavs, visible light imagery, thermal infrared data, lidar point clouds and synchronous positioning information to support the full section inspection of 72 km of the lede. A three-level cooperative track strategy and track adjustment based on backtracking search are introduced to maintain overlap, edge continuity and safe obstacle avoidance. Field test results show that the coverage integrity, data archipping and specification compliance verification reach 100%, a single 72 km inspection takes 5.2 hours, which is 32 times more efficient than manual inspection, reduces 80% of manpower, and can identify cracks and leakage risks with a width of not less than 1.8mm.
