基于混合特征的移动机器人图像匹配算法

中国机械工程 ›› 2015, Vol. 26 ›› Issue (9): 1137-1141.

基于混合特征的移动机器人图像匹配算法

陈卫东1,2;张兴家1;朱奇光1,2;陈颖1

1.燕山大学，秦皇岛，066004
2.河北省特种光纤与光纤传感重点实验室，秦皇岛，066004

出版日期:2015-05-10 发布日期:2015-05-08
基金资助:
国家自然科学基金资助项目(61201112,61172044)；河北省自然科学基金资助项目(F2013203250, F2012203169)；河北省普通高等学校青年拔尖人才计划资助项目(BJ2014056)；燕山大学青年教师自主研究计划资助项目(14LGA013)

An Image Matching Algorithm for Mobile Robot Localization Based on Hybrid Features

Chen Weidong1,2;Zhang Xingjia1;Zhu Qiguang1,2;Chen Ying1

1.Yanshan University,Qinhuangdao,Hebei,066004
2.The Key Laboratory for Special Fiber and Fiber Sensor of Hebei Province,Qinhuangdao,Hebei,066004

Online:2015-05-10 Published:2015-05-08
Supported by:
National Natural Science Foundation of China（No. 61201112,61172044）;Hebei Provincial Natural Science Foundation of China（No. F2013203250, F2012203169）;

摘要/Abstract

摘要：

针对移动机器人定位过程中视觉图像处理速度慢以及特征点提取与匹配实时性、准确性差的问题，提出了基于颜色矩的改进SIFT分级图像匹配算法。首先改进SIFT算法，扩大极值点检测范围;采用Sobel算子计算特征点的梯度方向和幅值;以向量夹角为准则度量SIFT特征相似性，提高SIFT特征提取与匹配的速度和精度。图像匹配时先采用颜色矩对环境图像序列进行相似性排序，改进SIFT特征，再与排序后图像依次进行精确匹配，分级匹配提高了移动机器人的定位速度和精度。实验结果表明：与原SIFT相比，改进SIFT提高了特征向量的显著性，误匹配率降低约9.2%，特征点数量减少约20%;分级匹配提高了图像匹配速度和精度，SIFT特征计算量减小60%，总体耗时缩短40%。达到移动机器人定位实时性和鲁棒性的目的。

关键词: 移动机器人, 图像匹配, 颜色矩, 改进SIFT

Abstract:

For the problems of slow image processing speed, poor real-time performance and accuracy of feature points extraction and matching in robot localization process, an improved SIFT hierarchical image matching algorithm was proposed based on color moment. Extreme point's detective range was extended, and the gradient directions and magnitude of feature points were calculated by using Sobel operator to improve SIFT performance; besides, the vector angle was taken as a criterion to measure SIFT features' similarity, thus the accuracy and speed of feature points extraction and matching were improved. In image matching process, image sequence was sorted by color moment, then the improved SIFT features were matched with these sorted images precisely.Experimental results show that the proposed algorithm is a real-time and robust method to mobile localization problem: compared with SIFT, the false matching rate of the proposed algorithm is lowered by 9.2%, and the number of feature points are reduced by 20%; hierarchical match improves the speed and accuracy of image matching, for the computation amount of SIFT features is reduced by 60% and the total time consume is reduced by 40%.

Key words: mobile robot, image matching, color moment, improved scale invariant feature transform(SIFT)

中图分类号:

TP391

陈卫东, 张兴家, 朱奇光, 陈颖. 基于混合特征的移动机器人图像匹配算法[J]. 中国机械工程, 2015, 26(9): 1137-1141.

Chen Weidong, Zhang Xingjia, Zhu Qiguang, Chen Ying. An Image Matching Algorithm for Mobile Robot Localization Based on Hybrid Features[J]. China Mechanical Engineering, 2015, 26(9): 1137-1141.

参考文献

[1]Davison A J, Murray D W. Simultaneous Localization and Map-building Using Active Vision[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(7): 865-880.
[2]章毓晋．图像处理和分析[M]．北京：清华大学出版社, 1999.
[3]王殿军. 双目视觉在移动机器人定位中的应用[J]. 中国机械工程, 2013, 24(9): 1155-1158.
Wang Dianjun.Application on Binocular Vision System of a Mobile Robot[J]. China Mechanical Engineering, 2013, 24(9): 1155-1158.
[4]张明路, 关柏青, 丁承君. 基于彩色视觉和模糊控制的移动机器人路径跟踪[J]. 中国机械工程, 2002, 13(8): 636-640.
Zhang Minglu, Guan Baiqing, Ding Chengjun. Mobile Robot’s Trajectory Following Based on Color Vision and Fuzzy Control[J]. China Mechanical Engineering, 2002, 13(8): 636-640.
[5]Aguilera G D, Gonzalvez R P, Lopez H D. A Robust and Hierarchical Approach for the Automatic Co-registration of Intensity and Visible Images[J]. Optics & Laser Technology, 2012, 44(6): 1915-1923.
[6]Lowe D G. Object Recognition from Local Scale-invariant Features[C]//Proceedings of the 7th IEEE International Conference on Computer Vision. Kerkyra, Greece, 1999: 1150-1157.
[7]Lowe D G. Distinctive Image Features from Scale-invariant Keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91-110.
[8]杨幸芳, 黄玉美, 韩旭, 等. 一种提高SIFT特征匹配效率的方法[J]. 中国机械工程, 2012, 23(11): 1297-1301.
Yang Xingfang, Huang Yumei, Han Xu, et al. A Method for Improving Matching Efficiency of SIFT Features[J]. China Mechanical Engineering, 2012, 23(11): 1297- 1301.
[9]Zahedi M, Salehi M S. Licenseplate Recognition System Based on SIFT Features[J]. Procedia Computer Science, 2011, 3: 998-1002.
[10]Josip K, Jakob V, Frédéric J. Modeling Spatial Layout with Fisher Vectors for Image Categorization[C]//Proceedings of the 13th IEEE International Conference on Computer Vision. Barcelona, 2011: 1487-1494.
[11]陈凤东, 洪炳镕. 一种基于图像的移动机器人泊位方法[J]. 机器人, 2010, 32(2): 166-170.
Chen Fengdong, Hong Bingrong. An Image-based Docking Method for Mobile Robot[J]. Robot, 2010, 32(2): 166-170.
[12]Valgren C, Lilienthal J A. Appearance-based Longterm Localization in Outdoor Environments[J]. Robotics and Autonomous Systems, 2010, 58(2): 149-156.
[13]曾峦, 顾大龙. 一种基于扇形区域分割的SIFT特征描述符[J]. 自动化学报, 2012, 38(9): 1513-1519.
Zeng Luan, Gu Dalong. A SIFT Feature Descriptor Based on Sector Area Partitioning[J]. Acta Automatica Sinica, 2012, 38(9): 1513-1519.
[14]Reza J A, Kossar J S, Fariborz M. A New Illumination Invariant Feature Based on SIFT Descriptor in Color Space[C]//Proceedings of the International Symposium on Robotics and Intelligent Sensors. Kuching, Sarawak, Malaysia, 2012: 305-311.
[15]Stricker M, Orengo M. Similarity of Color Images[J]. Proc. SPIE Storage and Retrieval for Image and Video Databases, 1995, 2420: 381-392.