| Abstract:To address the problems that visual features are susceptible to adverse environmental conditions, LiDAR has insufficient semantic representation, and continuous tracking of slender pipeline targets is difficult during low-altitude inspection of water intake pipelines, an unmanned aerial vehicle autonomous navigation method based on LiDAR and vision fusion was investigated; a local bird’s-eye view space was adopted as the unified multimodal representation carrier, and a fused representation of visual semantics, LiDAR geometry and inertial motion state was constructed; a gated state space model based on spatio-temporal observation confidence was proposed, in which an adaptive modality reliability gate was used to suppress degraded feature interference, while pipeline-topology-guided forward-priority scanning and local-neighborhood scanning were employed to enhance continuous centerline modeling; a local navigation cost function coupling centerline tracking, safe obstacle avoidance, observation angle and trajectory smoothness constraints was established; experimental results show that the proposed method achieved a pipeline tracking error of 0.17 m, a pose root mean square error of 0.13 m, an obstacle intersection over union of 88.2%, an obstacle avoidance success rate of 96.4%, and a single-frame inference time of 46.8 ms; the results indicate that the method can improve perception robustness, trajectory stability and onboard real-time navigation capability for unmanned aerial vehicle inspection of water intake pipelines in complex low-altitude environments. |
点此下载全文
次