摘要
为了降低跟踪系统状态维数、提高跟踪精度,从拉格朗日动力学原理出发,利用B样条形状空间理论,定义了新的动能、势能和衰减能量,推导了新的动态轮廓线跟踪动力学方程,并应用运动估计中的块匹配技术来寻找相邻帧中的对应点。结果表明:形状空间的维数远小于控制点数,增强了跟踪的稳定性;形状矩阵的正交化处理可以保证得到的动力学方程自然解耦,转化为若干个独立的单变量二阶振动系统;块匹配方法能够更准确地检测特征曲线。实验结果证明了方法的有效性。
In order to decrease system dimension and improve tracking precision, new kinematic energy, potential energy, dissipation energy are defined and a new dynamic contour ?tracking model is then derived u-sing B - spline shape space theory according to Lagrangian dynamics. Block match method is used to seek the correspondence of contour points between neighboring frames. The dimension of shape space is much smaller than the number of control points, so that the stability of tracking is enhanced. The multi - dimensional dynamical system can be transformed into several independent univariate second - order vibration systems because the normalization of shape matrix can ensure natural decoupling of dynamical equation. Block match method can detect feature curve more accurately. Experimental results prove the validity of our new method.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2003年第7期830-833,882,共5页
Journal of Harbin Institute of Technology
基金
国家自然科学基金(69775007
60075010)