基于局部一致性的特征匹配算法

Feature matching algorithm based on local consistency

针对特征匹配对尺度、光照变化敏感的问题,提出一种改进ORB特征提取方法,并采用基于局部一致性的方法进行匹配。首先采用改进ORB算法提取鲁棒性更强的特征点,并计算特征点的方向与描述子,接着采用暴力匹配进行粗匹配,最后根据运动平滑性的条件使用基于网格运动统计的方法剔除误匹配。实验结果表明所研究算法在尺度、光照等条件变化时匹配平均精度仍然大于95%,具有较好的匹配准确率和鲁棒性。

Aiming at the problem that feature matching is sensitive to scale and illumination changes, an improved ORB feature extraction method is proposed, and the method based on local consistency. Firstly, the improved ORB algorithm is used to extract the more robust feature points, and the direction and descriptor of the feature points are calculated. Then the brute force matching is used for rough matching. Finally, the method based on grid motion statistics is used to eliminate errors according to the condition of motion smoothness. The experimental results show that the matching accuracy of the proposed algorithm is still greater than 95% when the scale, illumination and other conditions change, which has better matching accuracy and robustness.