移动机器人定位图像匹配的快速局部特征算法34.doc

移动机器人定位图像匹配的快速局部特征算法* 刘萍萍1,2，赵宏伟1,2，臧雪柏1，王　慧1 （1　吉林大学计算机科学与技术学院　长春　130012;2　吉林大学符号计算与知识工程教育部重点实验室　长春　130012） 摘　要：利用定位图像的局部特征进行移动机器人导航和定位是近年来该领域的研究热点。针对经典局部特征提取和描述算法实时性不好，提出一种快速局部特征（fast local feature, FLF）的检测子和描述子算法。利用离散尺寸的均值滤波器估算LOG算子构建图像尺度空间，建立尺度和旋转不变的检测子。在尺度相关的邻域中利用规格化后的像素强度作为关联信息，建立局部邻域的描述子。模拟实验中，利用优选参数的FLF与经典SIFT算法对比，在识别率相当的条件下，FLF的运行时间是SIFT的1/3。在标准评估图片和移动机器人平台拍摄的定位图片构成的数据集下，FLF的匹配效果好于经典的SIFT和SURF算子。因而FLF是一种适合于实时应用的快速局部特征。 关键词：局部特征；图像匹配；尺度空间；物体识别 中图分类号：TP391.4　　文献标识码：A　　国家标准学科分类代码：510.2050 Fast local feature algorithm applied to mobile robot localization image matching Liu Pingping1,2, Zhao Hongwei1,2, Zang Xuebai1, Wang Hui1 （1 Department of Computer Science and Technology, Jilin University, Changchun 130012,China；2 Key Laboratory of Symbolic Computation and Knowledge Engineering, Ministry of Education, Jilin University, Changchun 130012, China) Abstract：Using local features of localization images to navigate and localize is a hotspot in recent mobile robot research. A novel, fast local feature in real time localization image matching is proposed in this article, called FLF (fast local feature). We use the mean filters in disperse sizes to approximate LOG to build the image scale space. Then a fast local character descriptor is proposed. Normalized pixel intensity is computed on characteristic scale and used as the contextual information to describe the local neighbor pixel intensity variations to the central point. In simulation experiment, the run time of FLF using optimal selection parameters is about 1/3 of that of classical SIFT arithmetic under comparative recognition rate. In another matching experiment, within which the datasets are composed of standard evaluation images and images taken by mobile robot platform, performance evaluations show that FLF generally outperforms SIFT and SURF operators in image scale, rotation, illumination and blur transformations. Therefore, FLF is suitable for real-time image matching applications. Key words：local fe