For the low accuracy and limit application feature which is becaused by force feedback devices and data gloves in the virtual sculpture in order to track cutting trajectory, machine vision was introduced into the field of virtual sculpture for the first time. According to the application requirements of this method, an experimental system and a tracking algorithm were designed. The algorithm has improved the detection of connected region in dynamic determining search area based on Kalman filtering, which makes the accurate positioning more efficiently. At last, the tracking speed and accuracy was validated to prove the method’s probability. 

virtual sculpture, binocular vision, target tracking, Kalman filtering, connected region detection

针对当前虚拟雕刻系统中采用力反馈器、数据手套进行成形轨迹跟踪而导致精确度低、应用局限性大的缺点，将机器视觉引入到虚拟雕刻领域中，设计了相应的实验系统和跟踪算法。该算法以Kalman滤波为基础，在动态搜索区域内对连通区域检测算法加以改进，以实现目标的高速精确定位。通过编程实验验证了轨迹跟踪的速度和精度，证明了所提出方法的可行性。

虚拟雕刻, 双目视觉, 目标跟踪, Kalman滤波, 连通区域检测