快速鲁棒的摄像机运动跟踪方法
|
摘要
针对普遍存在的摄像机运动实时跟踪效率低、目标易丢失等问题,提出一种快速鲁棒的跟踪方法。釆用FAST特征结合金字塔光流跟踪的方法,对目标特征点的提取和匹配进行优化,减少计算量;建立摄像机跟踪状态的检测与更新机制,实时检测有效跟踪点的数量并及时更新,保持摄像机运动跟踪的稳定性。实验结果表明,该方法有效减少了无标识增强现实中特征点提取与匹配的时间消耗,提高了摄像机运动跟踪的效率,具有良好的实时性,在遮挡及环境变化导致特征点数量减少时表现出较好的鲁棒性。
There usually exist such problems as low efficiency of real-time camera motion tracking and quickly losing of the tracked target points.To solve the problems,a fast robust tracking method was proposed.Two contributions were made.FAST features and pyramid optical flow were adopted to optimize the extracting and matching of target feature points,reducing computational cost.A new mechanism of detecting and updating of camera tracking state was adopted,which real-time detected the number of effective tracking points,and updated the tracking points once the number was too small,retaining the stability of camera motion tracking.Experimental results show the proposed method reduces time-consuming of extracting and matching of feature points in mark-less augmented reality,and improves camera motion tracking efficiency.Compared to other methods,the proposed method gets better real-time tracking ability with better robustness when the number of feature points drops as a result of object occluding and environment changing.
There usually exist such problems as low efficiency of real-time camera motion tracking and quickly losing of the tracked target points.To solve the problems,a fast robust tracking method was proposed.Two contributions were made.FAST features and pyramid optical flow were adopted to optimize the extracting and matching of target feature points,reducing computational cost.A new mechanism of detecting and updating of camera tracking state was adopted,which real-time detected the number of effective tracking points,and updated the tracking points once the number was too small,retaining the stability of camera motion tracking.Experimental results show the proposed method reduces time-consuming of extracting and matching of feature points in mark-less augmented reality,and improves camera motion tracking efficiency.Compared to other methods,the proposed method gets better real-time tracking ability with better robustness when the number of feature points drops as a result of object occluding and environment changing.
引文
[1]Klein A,De Assis GA.A markeless augmented reality tracking for enhancing the user interaction during virtual rehabilitation[C]//Virtual&Augmented Reality,2013:117-124.
[2]Shoaib H,Jaffry SW.A survey of augmented reality[C]//International Conference on Virtual and Augmented Reality,2015.
[3]Monji-Azad S,Kasaei S,Eftekhari-Moghadam AM.An efficient augmented reality method for sports scene visualization from single moving camera[C]//Electrical Engineering,2015:1064-1069.
[4]Hamidia M,Zenati-Henda N,Belghit H,et al.Markerless tracking using interest window for augmented reality applications[C]//International Conference on Multimedia Computing and Systems.IEEE,2014:20-25.
[5]Panchal PM,Panchal SR,Shah SK.A comparison of sift and surf[J].International Journal of Innovative Research in Computer&Communication Engineering,2013,1(2).
[6]Du M,Wang J,Li J,et al.Robot robust object recognition based on fast SURF feature matching[C]//Chinese Automation Congress,2013:581-586.
[7]Guerrero Pena M.A comparative study of three image matching algorithms:SIFT,SURF,and FAST[J].Dissertations&Theses-Gradworks,2011.
[8]ZHEN Chi,XIANG Zhiyu,LIU Jilin.Monocular vision odometry based on the fusion of optical flow an feature points matching[J].Journal of Zhejiang University(Engineering Science),2014,42(2):279-284(in Chinese).[郑驰,项志宇,刘济林.融合光流与特征点匹配的单目视觉里程计[J].浙江大学学报(工学版),2014,42(2):279-284.]
[9]Koniarski K.Augmented reality using optical flow[C]//Computer Science&Information Systems,2015:841-847.
[10]Patel D,Upadhyay S.Optical flow measurement using Lucas Kanade method[J].International Journal of Computer Applications,2013,61(10):6-10.
[2]Shoaib H,Jaffry SW.A survey of augmented reality[C]//International Conference on Virtual and Augmented Reality,2015.
[3]Monji-Azad S,Kasaei S,Eftekhari-Moghadam AM.An efficient augmented reality method for sports scene visualization from single moving camera[C]//Electrical Engineering,2015:1064-1069.
[4]Hamidia M,Zenati-Henda N,Belghit H,et al.Markerless tracking using interest window for augmented reality applications[C]//International Conference on Multimedia Computing and Systems.IEEE,2014:20-25.
[5]Panchal PM,Panchal SR,Shah SK.A comparison of sift and surf[J].International Journal of Innovative Research in Computer&Communication Engineering,2013,1(2).
[6]Du M,Wang J,Li J,et al.Robot robust object recognition based on fast SURF feature matching[C]//Chinese Automation Congress,2013:581-586.
[7]Guerrero Pena M.A comparative study of three image matching algorithms:SIFT,SURF,and FAST[J].Dissertations&Theses-Gradworks,2011.
[8]ZHEN Chi,XIANG Zhiyu,LIU Jilin.Monocular vision odometry based on the fusion of optical flow an feature points matching[J].Journal of Zhejiang University(Engineering Science),2014,42(2):279-284(in Chinese).[郑驰,项志宇,刘济林.融合光流与特征点匹配的单目视觉里程计[J].浙江大学学报(工学版),2014,42(2):279-284.]
[9]Koniarski K.Augmented reality using optical flow[C]//Computer Science&Information Systems,2015:841-847.
[10]Patel D,Upadhyay S.Optical flow measurement using Lucas Kanade method[J].International Journal of Computer Applications,2013,61(10):6-10.