基于SIFT和Hu特征融合的单目视觉识别算法研究

Research of monocular vision recognition algorithm based on SIFT and Hu feature fusion

研究了机器视觉技术在三维物体识别定位问题中的应用。利用Visual C++编程软件建立了摄像机标定界面,实现对摄像机进行快速标定的功能。提出了一种SIFT特征和Hu不变矩融合算法,该算法是一种融合了局部特征和全局特征的算法。其中全局特征反映了对三维物体图像的整体信息进行粗略匹配和定位,局部特征可以在全局特征中进行更准确特征匹配,该算法对伸缩、旋转和平移等有很好的抵抗能力。实验结果表明,该视觉算法可以有效解决三维物体匹配问题,并且有效提高了系统的识别速率和效率,满足物体识别的目的。

This paper researched machine vision technology on the application of the 3D object recognition and positioning problem. Visual C＋＋ pogramming software was set up for camera calibration interface, which realized function that camera was quickly calibrated. This paper put forward a kind of SIFT features and Hu invariant moment fusion algorithm, which was a kind of combined with local characteristics and global characteristics. The global features roughly matched and fixed position to reflect the overall informatiou of 3D object image, and local characteristics accurately matched feature based on global characteristics. The algorithm bas a good resistant ability for expansion, rotation and translation. The experimental results show that visual algorithm can effectively solve the matching problem of 3D objects and efficiency improves of system identification rate and efficiency to achieve the purpose of object reco~,~ition.