一种面向跨类别目标的无人机视觉主动跟踪方法及系统
By performing target detection and segmentation, scale standardization processing on monocular camera images, extracting depth and artificial features, and combining them with UAV state estimates, end-to-end control of the UAV is achieved using long short-term memory networks and deep networks. This solves the compatibility and generalization problems in cross-category target tracking, and improves tracking efficiency and practicality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TIANMUSHAN LABORATORY
- Filing Date
- 2026-03-23
- Publication Date
- 2026-07-17
AI Technical Summary
Existing UAV visual active tracking methods suffer from poor compatibility of computer vision tracking algorithms with edge devices and insufficient generalization of reinforcement learning methods when facing cross-category targets, resulting in a decline in tracking performance.
Image information is acquired using a monocular camera, and a mask is obtained through target detection and image segmentation. After scale standardization, depth features and artificial features are extracted. Combined with UAV state estimation, end-to-end control is performed using a long short-term memory network and a deep network, and three-dimensional continuous motion commands are output to achieve target tracking.
It improves the versatility and efficiency of UAVs in tracking cross-category targets, simplifies the system deployment process, and can output smooth and reasonable control commands without complex adjustments or retraining under different environments and targets.
Smart Images
Figure CN121884203B_ABST