基于特征点提取的轴承瑕疵工业在线检测  

Bearing defect online detection based on feature-point extraction

作　　者：柴先涛[1] 梁久祯[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《济南大学学报（自然科学版）》2015年第5期340-345,共6页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金(61170121)

摘　　要：基于ORB(oriented FAST and rotated BRIEF)图像特征点的配准方法实现差异性的比较,提取模板图像和待检测图像特征点,对特征点描述的集合进行匹配,通过RANSAC算法消除错误的匹配点,根据匹配对计算最优旋转角度,计算变换矩阵,通过仿射变换实现模板图像与检测图像的配准。在轴承图像上经过不同特征点提取算法,分析运行时间和图像配准的情况。结果表明,基于ORB提取特征点进行轴承瑕疵检测的方法,检测精度达96%,运行效率为67 ms。ORB（ oriented FAST and rotated BRIEF） algorithm is used to compare the difference between template images and the images to be detected. Firstly,feature points of the template image and the image to be detected are extracted and matched based on the set description of feature points to eliminate the false matching points by using the RANSAC algorithm; Secondly,the rotation angle and transformation matrix are calculated according to the optimal matching to realize image registration based on affine transformation. Experiments are performed to detect various bearing images by comparing different feature-point extraction algorithms. In addition,the running time and image registration situation are analyzed. Experimental results show that the proposed method has high accuracy which is more than 96% and operating efficiency of 67 ms.

关 键 词：特征点 图像配准 轴承防尘盖 表面缺陷 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]