Bayesian分类方法在微小型飞行器视觉导航中的应用  

Bayesian Classifier’s Application to Micro Aerial Vehicle’s Visual Navigation

作　　者：周宇[1] 黄显林[1] 介鸣[1] 宋卓异[1] 

机构地区：[1]哈尔滨工业大学控制理论与制导技术研究中心,黑龙江哈尔滨150001

出　　处：《航空学报》2007年第B08期136-140,共5页Acta Aeronautica et Astronautica Sinica

基　　金：国家自然科学基金(60475027)

摘　　要：在微小型飞行器的视觉导航中,图像分割的基本工作是将天地两部分分割开。这是形成整个视觉导航系统闭环控制的基础。而欲实现天地分割,则以特征线的提取为前提。本文分别从自底向上和自顶向下两种思路出发,利用广义Hough变换和基于多尺度最小方差的方法,进行了全局特征线的提取。然后,采用Bayesian分类的方法,实现了对微小型飞行器视觉导航至关重要的天地分割。通过具有典型意义的仿真实验,比较了两种方法所获得的结果,揭示了两种方法应用于微小型飞行器视觉导航的优点和缺点。In micro air vehicle＇s visual navigation, splitting the image into two parts （sky and ground） is a fundamental problem. This is the prerequisite of the visual navigation system＇s closed-loop control. However, the foundation of sky ground segmentation is the extraction of the characteristic line. By using the design philosophies of bottom-up and top-down respectively, both generalized Hough transform method and multi-scale least square based line extraction method are realized in this paper. Further, after the extraction of the characteristic line, a Bayesian classifier is established to segment the image into sky part and ground part respectively. Then, to give a comparative study, several typical simulations of these two methods are given. This study uncovers the advantages and disadvantages of these two methods in aircraft＇s visual navigation.

关 键 词：Bayesian分类 飞行器视觉导航 广义HOUGH变换 

分 类 号：V249.32[航空宇航科学与技术—飞行器设计]