本文为美国俄亥俄州立大学（作者：Ting Zhu）的硕士论文，共70页。

针对目标跟踪的算法有很多种，可分为概率和非概率两大类。在这两种情况下，现有的算法有时需要改进以满足特定应用的挑战，例如跟踪目标的突然运动、改变环境的光照条件、背景中具有相似外观的现有对象，一个好的算法必须对特定的应用具有鲁棒性，这通常会导致鲁棒性和效率之间的权衡。

在本课题的研究中，我们开发了一套能有效追踪食指指尖运动的系统，以取代电脑的滑鼠和键盘进行人机对话。我们称这种设置为手指鼠标。指尖用电子胶带标记为红色，背景是放置电脑的桌子表面。在分析人体指尖自然运动的基础上，提出了一种改进的粒子滤波算法先验运动模型。与现有的方法相比，本文提出的结合时间速度的高阶自回归模型能更准确有效地进行指尖跟踪。这项研究的结果将是非常有用的。除了提供健康人的替代品外，它还特别适用于不能机械移动鼠标，但使用指尖来表达移动光标意图的残疾人。

Various kinds of algorithms have been developed for object tracking, which can be divided into two categories, probabilistic and non-probabilistic, respectively. In either case, existing algorithms sometimes need to be improved to meet the challenges of a particular application, such as tracking abrupt motions of the target, changing lighting conditions of the environments, existing objects with similar appearance in the background, and etc. A good algorithm has to be robust for a particular application usually resulting in a trade-off between robustness and efficiency. In our research topic, we have developed a system to efficiently track the motion of the tip of the index finger, for the purpose of replacing the mouse and pad of a computer for HCI. We call this setup Finger Mouse implementation. The fingertip is marked by red using an electrical tape, and the background is the surface of the desk where the computer lays. We have developed a modified priori motion model for the particle filtering algorithm based on the analysis of natural motion of human fingertip movement. Our high-order autoregressive model combined with temporal velocity performs more accurately and efficiently for fingertip tracking, compared with the existing methods. The results of this research will be very useful. In addition to providing an alternative to healthy individuals, it is particularly suitable for disabled people who cannot mechanically move the mouse but use fingertip to express his/her intention of moving the cursor.

1.       引言

2. 指尖跟踪方法

3. 人体指尖的自然运动特性

4. 实验结果

5. 本文贡献及未来展望

更多精彩文章请关注公众号：