摘要
针对传统尺度不变特征变换(Scale Invariant Feature Transformation,SIFT)和加速鲁棒特征(Speed-Up Robust Feature,SURF)算法在视觉同步定位与建图(Simultaneous Localization And Mapping,SLAM)系统中耗时严重的问题,基于ORB(ORiented BRIEF(Binary Robust Independent Elementary Features))算法提出了一种改进的图像匹配算法。针对FAST(Features from Accelerated Segment Test)特征检测算子易受图像模糊和距离变化影响的缺点,建立了多尺度空间金字塔;针对BRIEF特征描述算子效率不高的问题,采用精简后的快速视网膜特征描述算子构建了特征向量;通过最邻近的交叉匹配对特征向量进行了提纯,采用顺序采样一致性算法剔除了错误匹配对。最后,通过与SIFT、SURF和ORB算法进行对比验证了改进算法的有效性。
To solve the long time-consuming problem of traditional feature matching algorithms based on SIFT( Scale Invariant Feature Transformation) and SURF( Speed-Up Robust Feature) in visual SLAM( Simultaneous Localization And Mapping) system,an improved image matching algorithm is proposed based on ORB( ORiented BRIEF( Binary Robust Independent Elementary Features)). Firstly,a multiscale space pyramid is constructed in accordance with the defect that the FAST( Features from Accelerated Segment Test) detectors are susceptible to image blurring and distance variance. Secondly,the feature vectors are built by the refined fast retina key point descriptors for inefficiency of BRIEF descriptors.Finally,the feature vectors are purified by the nearest cross matching pairs and the false matching pairs are eliminated by using progressive sample consensus algorithm. The validity of the proposed algorithm is verified by comparing with SIFT,SURF and ORB algorithms.
出处
《装甲兵工程学院学报》
2016年第6期82-88,共7页
Journal of Academy of Armored Force Engineering