基于自适应阈值的ORB特征点提取算法研究

Research on ORB feature point extraction algorithm based on adaptive threshold

在弱纹理场景下,针对ORB算法提取特征点过程中,固定阈值检测FAST角点可能会导致提取效果不佳进而影响匹配精度,提出了自适应阈值的ORB特征点提取算法,通过图像灰度差值和像素分布概率来计算图像对比度,根据对比度动态计算角点检测阈值。然后根据动态阈值算法实现特征点的提取,采用暴力匹配算法和快速最近邻接匹配(FLANN)两种匹配算法,在EuRoc数据集上分别对比了ORB算法、SIFT算法与该算法的特征点匹配精度和耗时。结果表明,在匹配精度上比ORB算法提升了26.6%,比SIFT算法提升了32.7%。

In weak texture scenes,in the process of extracting feature points with the ORB algorithm,fixed threshold detec⁃tion of FAST corner points may lead to poor extraction results and affect the matching accuracy.This paper proposes an adaptive threshold ORB feature point extraction algorithm,which uses image grayscale The difference value and pixel distribution probabil⁃ity are used to calculate the image contrast,and the corner detection threshold is dynamically calculated based on the contrast.Then the feature points were extracted based on the dynamic threshold algorithm,using two matching algorithms:brute force match⁃ing algorithm and fast nearest neighbor matching(FLANN).The feature point matching accuracy and accuracy of the ORB algo⁃rithm,SIFT algorithm and the algorithm in this paper were compared on the EuRoc data set.time consuming.The results show that the matching accuracy is 26.6%higher than the ORB algorithm and 32.7%higher than the SIFT algorithm.