一种改进的ORB算法在图像匹配中的应用

Application of an improved ORB algorithm in image matching

针对ORB算法提取的特征点阈值的选取存在人为干涉且对不同对比度的图像缺乏鲁棒性的问题,提出一种改进的ORB算法。本算法对FAST算法提取特征点进行了改进,首先对图像进行直方图均衡化实现对图像的增强,然后采用自适应阈值的方法,分别设置动态全局和局部阈值提取特征点,通过引用海森矩阵去除不稳定的边缘点。实验结果表明,该改进的算法能够实现特征点的精准定位,具有较强的抗噪能力,在实现图像配准应用中明显优于传统ORB算法。

The threshold of feature point in ORB algorithm exists problems of human intervention and lack of robustness in different contrast im-ages. Focusing on the problems, this article presents an improved algorithm of ORB. This algorithm improves FAST algorithm. Firstly, histo-gram equalization is used to realize the image enhancement. Then a metliod of adaptive threshold is introthreshold and a local threshold to extract the feature points respectively. At last, unstable feature poinExperiment shows that the improved algorithm can realize the accurate positioning of feature points, and ty , it is superior to the traditional algorithms of the ORB in the realization of image matching.