变圆域罗盘特征图像匹配

Image matching of variable circular domain compass features

针对现有图像匹配算法特征算子计算复杂度高,关键点描述过程中点对选取存在不确定性等问题,提出了变圆域罗盘特征匹配(VCDCM)方法。该方法首先利用罗盘检测法对图像进行关键点检测,采用变圆模式接受域选取理想点对,根据接受域内点对之间的距离将点对分为长点对集和短点对集;然后用长点对集描述关键点方向,短点对集构建关键点描述符。最后采用Hamming距离代替传统的欧式距离进行匹配,并采用随机抽样一致(RANSAC)方法精炼匹配点以避免由于噪声和物体位置移动等原因产生的误匹配。从鲁棒性和实时性两个方面对本文提出的方法与尺度不变特征变换(SIFT)和二元加速鲁棒特征(BRIEF)方法进行了对比试验分析,实验结果表明,本文提出的方法具有匹配速度快、准确性高、稳定性好等特点。

As existing image matching algorithms show the problems of high computational complexity and uncertainty in point-pair selection, a Variable Circular Domain Compass Matching （VCDCNI） al- gorithm was proposed. After key-points being detected by four compasses, the variable circular re- ceiving domain was used to choose the ideal point-pairs. Then, the point-pair set was divided into two subsets according to the distance of point-pairs. The subset of long-distance pairs was used to describe the direction of a key-point and that of short-distance pairs was used to build the descriptor of key- point. Finally, key-points were matched by Hamming distance instead of traditional Euclidean dis- tance, while the match points were filtered with Random Sample Consensus （RANSAC） algorithm to avoid mismatches caused by the noise and moving objects. Comparative experiments between Scale In- variant Feature Transform （SIFT） and Binary Robust Independent Elementary Features （BRIEF） al- gorithms were performed on the robustness and efficiency. The experimental results show that the proposed algorithm is faster with high accuracy and stability.