融合彩色不变量和SURB检测的特征点匹配算法  

Feature Point Matching Algorithm for Fusion of Color Invariants and SURB Detection

作　　者：陈树[1] 杨天 孙顺远[1] Chen Shu;Yang Tian;Sun Shunyuan(School of Internet of Things Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China)

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《激光与光电子学进展》2018年第5期138-145,共8页Laser & Optoelectronics Progress

摘　　要：为了解决传统匹配算法运行效率低、匹配精度低等问题,在快速特征点提取和描述(ORB)算法的基础上提出一种融合彩色不变量和基于加速稳健特征和对象请求代理(SURB)检测的优化匹配方法。首先,在同一复杂场景下,通过彩色空间变换提取两幅图片的彩色不变量信息;然后,采用SURB算法提取彩色信息中具有尺度不变性的特征点,构建ORB算术描述子;最后,在K近邻算法分类和整理的基础上,采用评价函数和随机抽样一致性算法去除误匹配点对。基于标准图库ALOI及多组实际图像匹配的结果表明,对于复杂环境下的图像,本文算法不仅具有较高的匹配精度,而且大幅缩短了匹配时间。Aiming at the low operating efficiency and low matching accuracy problems in image matching, we propose an optimized matching method based on the color inwtriants and SURB [speeded up robust features-oriented FAST and rotated BRIEF （SURF-ORB）] detection on the basis of ORB matching algorithm. First of all, we obtain the color invarianee of two images in the same complex scene through the color space transformation. Then, we use SURB algorithm to extract the feature points with the scale invariance and build an ORB arithmetic descriptor in chromatic information. At last, on the basis of classification and disposal of K nearest neighbor algorithm, we use the cost function and random sample consensus algorithm to remove the mismatching pairs. Results of standard graphics library ALOI and many other real images indicate that the proposed algorithm improves the image matching accuracy and shortens the image matching time in the complex environment.

关 键 词：图像处理 特征点匹配 彩色不变量 信息熵 评价函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]