Abstract: A machine vision system has been developed for solution scheme for automatic landing of an unmanned helicopter on deck.For recognition known object，the system first captures the image，and then threshold，median filter，segment and label the helipad in image，final the system calculates the Affine Moment Invariants（AMI）.When the error of the distance of AMI and standard value from database is about 10%，the system finds the helipad.In experimentation，the estimate error of the helipad center is small 3 pixels each axis and the estimate error of the azimuth is small 9 degree.

Key words: unmanned helicopter, automatic landing, affine moment invariants, machine vision

摘要： 根据无人直升机自主着舰的任务特点，构造了用于无人直升机自主着舰的视觉系统，选用仿射不变矩作为该视觉系统目标识别的特征和信息提取的手段，仿真实验表明，开发的算法能够在纵摇角小于15°的情况下正确的识别出目标，对目标中心位置的估计偏差在每个坐标轴向小于3个像素，方位角估计偏差小于9°。整套算法的鲁棒性和实时性满足无人直升机自主着舰的需求。

关键词: 无人直升机, 自主着舰, 仿射不变矩, 机器视觉