基于改进ORB算法的特征检测与匹配研究

Research on Feature Detection and Matching Based on Improved ORB Algorithm

针对传统ORB算法中阙值选取不具有普适性,易受光照影响,纹理信息提取能力弱且特征点易产生富集情况的缺陷,提出改进的ORB特征提取与匹配算法。首先以待检测点为中心进行区域划分,其次利用自适应阙值判断待检测点,并使用自适应范围的非极大抑制算法解决特征点富集问题,提高均匀性。然后使用各项异性扩散滤波提高纹理信息提取能力,最后结合双边滤波改进描述子点对,结合PROSAC算法消除误匹配。实验结果表明,改进后的算法对光照有较强的适应能力,提高了纹理特征提取的能力,特征点分布均匀,改善了富集效应,光照条件下的匹配精度得到提高,具有更好的适应能力。

This paper proposes an improvement to the ORB feature extraction and matching algorithm to address the shortcom-ings of the traditional version,which is susceptible to errors due to its reliance on a fixed queue value,sensitivity to lighting condi-tions,limited ability to extract texture information,limited ability to extract texture information,and a tendency for feature point en-richment.Firstly,the region is divided with the point to be detected as the centre.Secondly,an adaptive threshold is used to judge the detection of the point,and an adaptive range non-maximum suppression algorithm is applied to address the feature point enrich-ment issue,thereby enhancing uniformity.Thirdly,the texture information extraction ability is improved using various anisotropic diffusion filtering.Finally,the descriptor pairs are improved by combining bilateral filtering,and the mismatch is eliminated by the PROSAC algorithm.The experimental results demonstrate that the enhanced algorithm exhibits a robust capacity to adapt to varying lighting conditions,enhances the capability to extract texture features,ensures a more uniform distribution of feature points,im-proves the enrichment effect,and enhances the accuracy of matching under light conditions.It displays superior adaptive capabili-ties.