基于视觉模糊的LBP鲁棒特征提取与匹配  

LBP robust feature extraction and matching based on visual blur

作　　者：曹亚媛 郭秀才[1] 程勇[1] CAO Ya-yuan;GUO Xiu-cai;CHENG Yong(School of Electrical and Control Engineering,Xi′an University of Science and Technology,Xi′an,Shaanxi 710054,China)

机构地区：[1]西安科技大学电气与控制工程学院,陕西西安710054

出　　处：《光电子．激光》2021年第4期361-372,共12页Journal of Optoelectronics·Laser

摘　　要：目的针对目前模糊图像特征提取与匹配方面,存在特征提取困难、匹配率低、抗噪以及抗尺度变化能力弱的缺陷。方法提出一种基于SIFT算法与改进的中心对称局部二值模式相结合的精准、特征识别率高的匹配算法。首先采用SIFT进行特征的提取,生成多维的描述子,其次采用本文改进的中心对称局部二值模式对高维特征描述子进行降维处理,并采用局部特征区域对降维后的描述子进行特征检测,并生成纹理特征图像以及信息分布直方图,对特征区域的特征点进行信息量统计,并设置检测阈值。提取符合特征信息要求的特征点,并依据Hausdorff距离算法实现图像粗匹配,最后采用RANSAC算法进行误差匹配的剔除来改善匹配的精度和鲁棒性。结果测试结果表明,本文所建议的算法是有效的,它不仅具有良好的模糊图像分辨能力和抗尺度变化特性,而且具有较强的噪声抑制能力和抗光照变化能力。结论本文提出的基于视觉模糊的鲁棒特征匹配算法,不仅考虑到传统特征匹配算法的优缺点,也提出了算法改进的新思路,而且较SIFT算法以及LBP算法稳定性和准确度有了明显的提高。Objective In view of the current feature extraction and matching of fuzzy images,there are defects such as difficulty in feature extraction,low matching rate,weak anti-noise and anti-scale changes.Methods A matching algorithm with high accuracy and high feature recognition rate based on the combination of SIFT algorithm and improved center symmetric local binary pattern is proposed.First,SIFT is used for feature extraction to generate multidimensional descriptors.Second,the improved center symmetric local binary mode is used to reduce the dimension of high-dimensional feature descriptors,and local feature regions are used to feature the reduced-dimensional descriptors.Detect and generate texture feature images and information distribution histograms,perform information statistics on feature points in feature areas,and set detection thresholds.Extract feature points that meet the requirements of feature information,and implement rough image matching based on Hausdorff distance algorithm,and finally use RANSAC algorithm to eliminate error matching to improve the accuracy and robustness of matching.Results The test results show that the algorithm proposed in this paper is effective.It not only has good fuzzy image resolution and anti-scale change characteristics,but also has strong noise suppression and anti-light changes.Conclusion The robust feature matching algorithm based on visual blur proposed in this paper not only considers the advantages and disadvantages of traditional feature matching algorithms,but also proposes new ideas for algorithm improvement,and has obvious stability and accuracy compared with SIFT and LBP algorithms improve.

关 键 词：视觉模糊 SIFT算法 中心对称局部二值模式 Hausdorff距离算法 RANSAC算法 特征检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]