基于一种层叠分类器的无人直升机路标识别算法  

Unmanned Helicopter Landmark Recognition Algorithm Based on a Cascade Classifier

作　　者：陈林利[1] 王慧[1] 任沁源[1] 

机构地区：[1]浙江大学信息科学与工程学院工业控制技术研究所,浙江杭州310027

出　　处：《江南大学学报（自然科学版）》2008年第3期299-303,共5页Joural of Jiangnan University (Natural Science Edition)　

摘　　要：针对复杂的野外环境造成的图像表象变化给无人直升机飞行过程中路标识别带来的困难,分析了常规神经网络分类器和Hu不变矩分类算法的优、缺点,提出一种利用二级神经网络结合不变矩等特征构成的层叠分类器的路标识别算法,给出了算法的原理和结构.实验结果表明,此算法不仅能获得更好的识别效果,而且能提供路标的方位信息,可作为无人直升机视觉导航系统和进行姿态估计时的参考信息.Recently, unmanned helicopter landmark recognition has been studied more and more widely and deeply. In this paper, a cascade classifier, composed of two different neural networks, is proposed to detect a certain landmark which is difficult to recognize because of the complicated entironment. The principle and framework of the classifier is addressed on the basis of the neural networks theory and the Hu invariable moments theory. The test result is also presented, which shows that the cascade classifier can not only recognize the landmark more effectively, but also provide its orientation as a reference information for unmanned helicopter vision-based navigation and its pose estimation.

关 键 词：无人直升机 路标识别 神经网络 不变矩 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]