无人机视觉着舰导航鲁棒角点精确检测算法  

Robust corner precise detection algorithm for visual landing navigation of UAV

作　　者：郝帅[1] 程咏梅[1] 马旭[1] 王涛[2] 赵建涛[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072 [2]中国航空工业集团西安飞行自动控制研究所重点实验室,陕西西安710065

出　　处：《系统工程与电子技术》2013年第6期1262-1267,共6页Systems Engineering and Electronics

基　　金：航空科学基金(20100853010)资助课题

摘　　要：针对舰载无人直升机着舰时拍摄的图像存在大尺度、角度畸变,使得合作目标角点难以检测的问题,提出了一种鲁棒角点精确检测算法。设计了一种彩色非对称合作目标,通过计算合作目标的最小周长多边形(minimum perimeter polygon,MPP)来获取轮廓上的角点。针对这些角点存在伪角点及未排序问题,给出了斜率约束、距离约束和方向约束策略。最后,针对亚像素Harris角点检测算法精度高但未排序、三重约束的MPP角点检测精度低但已排序的问题,制定了这两种角点检测方法的融合策略。实验结果表明,所提算法可以有效解决合作目标图像发生大角度畸变以及一定尺度畸变范围内的角点检测。A robust corner precise detection algorithm is proposed to solve the problem that the images have a se rious distortion of scale and angle when the carrier-based unmanned helicopter lands. A cooperation with distinct color contrast and asymmetric is designed. And corners on the cooperative object can be gotten through calculating the mini mum perimeter polygon （MPP） of the cooperative object. To solve the problem of removing false corners and corners out of order, a strategy which concludes slope constrain, distance constrain and orientation constrain is proposed. Fi nally, to solve the problem of the sub pixel Harris corner detection algorithm with high precision but out of order and the MPP corner detection algorithm based on triple constraints with low precision hut ordered, a fusion strategy is pro posed. Experiment results show that the proposed fusion algorithm can effectively solve the problem of corner detection when cooperation has a serious distortion of angle and a certain range of scale distortion.

关 键 词：视觉着舰 三重约束 最小周长多边形 融合策略 亚像素 

分 类 号：TP217.3[自动化与计算机技术—检测技术与自动化装置]