基于固定翼飞机前视近红外图像的实时跑道检测

为实时、精确地提取固定翼飞机前视图像中的跑道区域,提出了一种惯性辅助的近红外图像跑道检测算法。首先利用惯性导航信息和跑道角点的大地坐标,生成虚拟跑道轮廓,并估计跑道感兴趣区域(Region of Interest,ROI);然后对ROI区域进行细节增强,从中提取直线段;在虚拟跑道轮廓的辅助下拟合跑道边缘线,提取跑道特征。经真实数据验证,所提法可以从机载前视红外图像中实时、精确地检测出跑道特征,在33 frame/s的帧率下检测准确率达97%。

Real-time runway detection for forward-looking near-infrared images of fix-wing aircraft

To extract runway regions of forward-looking near-infrared images for fixed-wing aircraft accurately and timely,an inertial aided runway detection algorithm is proposed.Firstly,by using the inertial navigation information and the geodetic coordinates of runway corners,the virtual runway contour is generated and the runway region of interest(ROI) is estimated.Then,the texture details of the ROI are enhanced to extract the line segments.With the help of the virtual runway contour,the runway edge lines are fitted and the runway features are extracted.Experiments on real flight data demonstrate that accurate runway feature detection in real-time can be achieved,and the detection accuracy of the algorithm is 97% at the frame rate of 33 frame/s.