【作者】 赵昊昱；

【导师】 李红；

【作者基本信息】 华中科技大学 ， 计算数学， 2006， 硕士

【摘要】 计算机视觉由于其经济、无源、信息丰富等特点,成为无人机自主着陆中不可缺少的重要信息源,利用视觉信息对飞机着陆进行导航已经成为一个崭新的研究课题。本文以国家自然科学基金项目”基于视觉的无人飞行器自主着陆导引信息实时提取与转换”为背景,以机载摄像机获得的飞机着陆过程中的机场序列图像为研究对象,研究利用计算机视觉的理论,通过数字图像处理方法以及代数与几何知识,实时提取飞行参数,为飞机自主着陆提供导航信息。本文主要在以下四个方面提出了一些新的方法和尝试。首先,在特征提取阶段,文中提出了一种局部使用方向小波变换的方法,该方法提取出的特征具有信息完整性、准确性以及抗噪性等特点。其次,在直线检测阶段,文中提出一种基于线性函数与二次函数相结合的预测方法,将其与Hough直线检测方法相结合,大大提高了运算的速度和检测的精度。在此基础上,文中提出一种计算飞机飞行参数的新方法,它克服了传统方法中利用点对应关系而难于提取特征点的问题,也不需要人为设置着陆标识,只需对两条跑道边线和一条着陆线进行检测,这些线具有明显的可视性和直线性,由此可直接算出飞机的三个姿态角;也可以利用地平线和跑道边线实现上述目的。若跑道的宽度是已知的,进而可以算出飞机的位置矢量。最后,本文尝试了一种直接利用图像特征,对飞机的飞行状况进行控制的方法。在控制调整过程中,用三次样条插值方法,利用三次样条的光滑性可以使飞机整个调整过程能够平稳完成。理论分析和实验结果表明,基于视觉的飞机自主着陆导航是可行的,而且具有一定的实时性和较高的计算精度,适合为飞机实时着陆进行导航。更多还原

【Abstract】 Computer vision is of the characters of economy, passivity and information abundance.So it has became an indispensable information source in the landing of the unmanned air-craft. This article bases on the national natural and science fond projection, studies theapplication of the computer vision in the navigation for the unmanned aircraft landing onthe airport. In this article, we process the image sequence of the runway which is got froman onboard camera, estimate the attitude and position of the aircraft.At first, in the character detection, this article takes use of the directional wavelets. Thecharacters based on this method are of integrate information, accuracy and antinoise.secondly, in the lines extraction, via a technique integrating the linear and quadraticpredictors, Hough transform is utilized to detect the lines. This improves the computingspeed and the accuracy.Then we propose a new extraction method of vision-based navigation information foraircraft autonomous landing. It avoids the difficulty of feature points extraction in the tradi-tional method using points correspondence relation. It also doesn’t need to set some arti?ciallandmark. It extracts the two edge lines and the threshold line on the runway only. Theselines are obviously visual and straight Then the attitude vector of aircraft is directly calcu-lated. And the position vector of aircraft is computed too if the runway width is known. Acomplete drive of the algorithm is given based on imaging geometry.In the adjustment, we try to utilize the character of the image to control the ?ying situ-ation. the Cubic Spline interpolation is used so that the adjustment is smooth and ef?cient.The theoretical analysis and the experiment result show that the extraction and conver-sion of vision-based navigation information for aircraft autonomous landing are feasible.It is very simple and has a high precision, so very suitable for a real-time navigation foraircraft landing.更多还原