基于ORB的目标识别与角度测量方法  

Object recognition and angle measurement method based on ORB

作　　者：徐洪 吴斌[1] 陈春梅[1] XU Hong;WU Bin;CHEN Chunmei(School of Information Engineering,Southwest University of Science and Technology,Mianyang 621010,China)

机构地区：[1]西南科技大学信息工程学院

出　　处：《传感器与微系统》2019年第12期143-147,共5页Transducer and Microsystem Technologies

摘　　要：针对利用图像配准识别目标及测量目标旋转角度的问题,提出了基于ORB的快速角度测量方法。在特征匹配识别目标的同时,利用ORB特征的旋转不变性,根据匹配的特征点的主方向角度差值,得到相匹配点的旋转角度。然后对角度数据进行迭代自组织聚类(ISODATA)分析,剔除错误角度数据,并以剩余角度数据的均值作为目标的旋转角度。实验结果表明:所提方法能够对不同尺度、旋转、遮挡的目标进行快速识别定位及角度测量,角度误差在0. 3°左右,消耗的时间在0. 6 s左右,相对于传统算法在测量精度及时间复杂度方面均有了较大的提高。Aiming at the problem of using image registration to identify the target and measure the target rotation angle,a fast angle measurement method based on ORB is proposed. While the feature matching identifies the target,using the rotation invariance of the ORB feature,rotation angle of the matching point is obtained according to angle difference of the main direction of the matched feature points. Then,the angle data is subjected to iterative self-organizing clustering(ISODATA) analysis,the wrong angle data is eliminated,and the average value of the remaining angle data is taken as the target rotation angle. Experimental results show that the proposed method can quickly identify,locate and can measure angle on different scales,rotation and occluded target. The angle error is about 0. 3°,and the consumed time is about 0. 6 s. The measurement precision and time complexity are greatly improved compared with traditional algorithm.

关 键 词：ORB特征 特征匹配 特征点主方向 聚类分析 角度测量 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]