视觉着舰中非合作结构化特征匹配算法

Non-cooperative Structural Feature Matching Algorithm in Visual Landing

视觉全自主着舰具有独立性强、系统结构简单以及测量信息丰富等优点,特别是非合作视觉着舰意义更为重大,但也对参考特征提取提出了更高的要求。为能够实现航母图像中参考定位非合作特征点的稳定、高精度匹配,本文提出一种基于线段检测子(Line segment detector,LSD)直线识别的结构性特征检测与匹配方法,实现对航母表面各种结构、标志角点的准确识别,可克服目前有向FAST与旋转BRIEF(Oriented FAST and rotated BRIEF,ORB)及加速鲁棒特征(Speeded up robust feature,SURF)等图像特征无法定位指定结构特征的问题,有效提高视觉定位精度。

Compared with other landing methods,visual autonomous landing has the advantages of obviously independent and simple system structure,and rich measurement information.The non-cooperative visual landing has great significance,but it also has higher requirements for the accuracy of feature extraction.In order to realize the stable and high-precision matching of non-cooperative feature points of reference location in the aircraft carrier image,this paper proposes a structural feature detecting and matching method based on straight line detection of line segment detector(LSD)to achieve accurate recognition of structural corner points on the aircraft carrier surface.Experiments show that compared with oriented FAST and rotated BRIEF(ORB),speeded up robust feature(SURF)and other image features,the proposed method can locate specified structural features more accurately and thus can improve the accuracy of visual positioning effectively.