电动VTOL飞行器双目立体视觉导航方法

Binocular stereo vision navigation for electric VTOL aircraft

研究了在未知的、动态的室内走廊环境中,采用双目立体视觉引导电动VTOL(Vertical Take-Off and Landing)飞行器安全飞行的方法.使用安装在飞机上的两个微型摄像头从不同的位置获取图像,由双目立体视觉理论恢复其周围环境特征点的三维坐标.采用角点匹配方法计算视差,实现无人机在走廊中的横向坐标定位.采用区域灰度相关算法进行立体匹配获取视差图,从视差图上检测出障碍物,并给出避障导航点.初步实验验证表明,该方法可行性较高,可以作为进一步研究的基础.

A stereo vision-based navigation method for the electric vertical take-off and landing（VTOL） aircraft to wander safely through an unknown dynamic indoor corridor was researched.The aircraft uses two cameras to obtain images from different locations,and then the theory of binocular stereo vision was used to restore the 3-D coordinates of the feature points in front of the aircraft.Corner matching method was applied to calculate the disparity of the detected corners on the wall.Thus,the horizontal position of unmanned aerial vehicle（UAV） in the corridor was revealed.The area-based stereo matching algorithm was presented to get the dense disparity map of the original image pair,then obstacles were extracted from the disparity map and the navigation points were generated for the UAV to avoid obstacles.Preliminary experiment shows that the method is feasible and can be used as a basis for further research.