Abstract: This paper designs a vision-based system to enable an unmanned helicopter to land on a moving platform.Based on affine moment invariants，fuzzy recognition，Kalman filter and a PI controller，the helicopter can identify a target，track it，and land on it while the target is in motion even oscillation.A simulation system，including an unmanned helicopter and a moveable platform with a six degrees of freedom deck，has been developed for test the ideas.Simulation results validate the algorithm.In simulation this algorithm has successfully isolated the influence of helicopter’s oscillation and found the best threshold between performance and precision.The results show that the algorithm is robust，the recognition algorithm can perform exactly at 97% and the tracker can track target in two-dimensions with a biggest angle error at 22.26°.

Key words: unmanned helicopter, autonomous landing, Kalman filter, affine moment, fuzzy recognition

摘要： 设计了一个视觉系统来实现无人直升降落在移动目标上。基于仿射不变矩、模糊识别，卡尔曼滤波和PI控制技术，直升机能够完成识别、跟踪目标的任务，并能够在目标运动的状态下完成降落。一套软件仿真系统，其中包含无人直升机和一个可移动的平台被用来测试此设计。仿真结果验证了该算法。在仿真中，成功地隔离了直升机的振动对算法的影响，并成功地处理了性能与精确性的矛盾。结果显示算法具有鲁棒性，能够以97%的概率正确的识别目标，并能在目标进行二维平面运动时进行精确的跟踪，跟踪时直升机的航向与目标的运动方向最大误差为22.26°。

关键词: 无人直升机, 自主降落, 卡尔曼滤波, 仿射矩, 模糊识别