欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
改进了增强现实中跟踪直线边缘对象位姿的方法。首先用扩展卡尔曼滤波器预测对象位姿并限制隐式边缘搜索的区域;然后利用M估计子与权重直线拟合得到图像中直线边缘的方程系数;最后利用平面的标准正交基构建关于对象位姿的目标函数,并用奇异值分解法加以求解。实验表明本文位姿跟踪方法的旋转角最大的平均偏差为0.29°,绝对偏差为1.90°,跟踪速度大约为10帧/s,综合性能明显优于基于LM的跟踪算法。
英文摘要：
The algorithm of pose tracking for object with straight lines in augmented reality is improved.Firstly the extended Kalman filter is used to predict object pose and the implicit searching regionis is restricted.Then,M-estimator and weight line fitting are used to obtain edge equation coefficients in image.Finally,the objective function about object pose is constructed using canonical basis of plane which is solved by singular value decomposition.Experiment shows that maximum mean deviation is 0.29° and absolute deviation 1.90° for rotation angles,and tracking rate is about 10 frame/s,so its comprehensive performance exceeds obviously the LM-based tracking algorithm.
同期刊论文项目
同项目期刊论文
Application of a state space modeling technique to system prognostics based on a health index for co
Prognostics Uncertainty Reduction by Fusing On-line Monitoring Data Based on a State-Space-Based Deg
Experimental study on engine gas-path components fault monitoring using exhaust gas electrostatic si
期刊信息
位置：