小型无人机着陆阶段视觉导航算法的验证

Verification on visual measurement algorithms for UAV’s landing task

文中以无人机视觉着陆为研究目标,采用视觉测量的方式引导无人机着陆。文中搭建了无人机平台,设计了着陆标识,通过机载相机拍摄包含着陆标识的图像序列引导着陆。机载处理器先通过局部二值模式与支持向量机得到初步的感兴趣区域,然后利用颜色信息进行特征提取和对应点匹配,最后利用鲁棒的透视n点定位方法测量,得到无人机与标识之间的精确位置。文中将上述算法放到无人机平台中进行验证,验证了算法的有效性。

This paper mainly focuses on vision based UAV autonomous landing task,which uses vision to guide the UAV to land.First,a UAV platform is built up with a camera capturing sequence images of landing mark.The on⁃board processor obtains a preliminary region of interest using Local Binary Patterns(LBP)classifying by Support Vector Machine(SVM).Then color segmentation is used to get the feature of the landing mark.Then Point Matching procedure is processed.Robust Perspective⁃n⁃Point(RPnP)is used to get the relative position and pose between the UAV and the landing mark.Finally,these algorithms are integrated into a UAV platform to verify the feasibility and effectiveness.