基于CenSurE-star和LDB的被测目标图像匹配算法

Target image matching algorithm based on CenSurE-star and LDB

摘要:

针对传统点特征匹配方法计算量大、匹配速度慢的问题,给出了一种基于CenSurE-star和LDB的图像匹配算法,以用于在视觉检测中对被测目标图像进行快速匹配。该算法首先通过调整滤波器尺寸从而快速检测被测目标图像中不同尺度的CenSurE-star特征点,然后采用LDB方法对特征点结合其邻域进行描述,以描述符汉明距离为标准衡量图像特征点间的相似度并进行相应筛选,最终结合RANSAC剔除剩余的误匹配点对,实现了图像间准确匹配。实验研究表明,在关于光照、噪声和模糊变化的三组被测目标图像匹配中相较SIFT、SURF等常见算法,该算法不仅显著提升匹配速度,而且保证了较高的匹配准确率。

Abstract:

In order to solve the problems of large amount of calculation and slow matching speed in the traditional point feature matching methods, an image matching algorithm based on CenSurE-star and LDB is proposed, which can be used to quickly match the target images in visual detection. Firstly, the size of the filter is adjusted to fast detect the CenSurE-star feature points of different scales in the target image. Then, the LDB method is used to describe the feature points combined with their neighborhood. The similarity between the image feature points is measured by the Hamming distance between descriptors, and the corresponding filtering is carried out. Finally, RANSAC is used to eliminate the remaining mismatches and achieve the accurate matching between the images. Experimental research shows that compared with common algorithms such as SIFT and SURF in the matching of the three sets of target images with regard to illumination, noise and blur changes, this algorithm not only improves the matching speed, but also ensures high matching accuracy.