基于主元分析的微小型飞行器视觉导航

Micro Air Vehicle Visual Navigation Based on Principal Component Analysis

作为一种新兴的无人机,微小型飞行器(MAV)近年来广受关注。研究微小型飞行器导航技术的关键是研究视觉导航技术在微小型飞行器中的应用。针对微小型飞行器视觉导航中的运算量大而导致算法实时性差的问题,给出了一个新的计算框架。利用主元分析技术(PCA),在最小方差意义下得到了图像的简化表示。在此简化的基础上,进行了图像特征的提取。对真实图像的实验结果表明,利用主元分析处理后的图像进行导航特征的提取,其实时性优于RGB三通道求和取平均的方法。

Micro air vehicle （MAV） is a mini-sized member of the family of unmanned aerial vehicles which receives widely attention recently. The key point in studying navigation technology of MAV is to explore the application of the visual navigation. The paper presents a new computational framework which is designed to improve the realtime-ness of the visual navigation system in MAV due to large volume of calculation. Using principal component analysis （PCA）, a simplified representation of the image is obtained in the sense of minimum squared errors. Furthermore the navigational features are further calculated based on this new representation. Simulation results of the real images show that this method is of better realtime-ness than the averaging approach of RGB channels respectively.