摘要
针对基于标识物的增强现实跟踪注册方法对复杂环境的适应能力和鲁棒性的不足,提出一种用SURF实现标识物跟踪注册的改进算法。用SURF对平面标识物的特征点进行检测与描述,采用K-means算法对检测出的标识物特征点集合进行聚类分析得到其聚类中心,结合对视频图像中标识物的阈值分割与连通域分析,以聚类中心构建标识物的SURF特征点坐标系,通过矩阵变换实现标识物的跟踪注册。实验结果表明,该算法的增强现实系统具备较好的鲁棒性、稳定性和实时性。
Aiming at the lack of adaptive faculty and the robustness of Augmented Reality(AR) of tracking and registration based on marker,this paper presents an improved algorithm of the marker’s tracking registration using SURF.The algorithm detects and describes the feature points of a plane marker through SURF,gets the clustering centers of the detected feature points of the marker by K-Means,combines threshold segmentation with connected component analysis to construct coordinates of SURF feature points,and realizes the tracking registration of the marker by means of matrix transformation.Experimental results show the robustness,stability and real-time of the AR system based on the improved algorithm.
出处
《计算机工程》
CAS
CSCD
北大核心
2010年第13期254-256,259,共4页
Computer Engineering
基金
安徽省高等学校省级优秀青年人才基金资助项目(2009SQRZ089)
安徽省高等学校省级自然科学研究基金资助重点项目(KJ2010A304)
淮北师范大学教研基金资助项目
关键词
增强现实
SURF算子
跟踪
注册
Augmented Reality(AR)
SURF operator
tracking
registration