中文摘要：

相对于需要第三方辅助设备的系统，基于视觉的手势跟踪更方便自然。单目的手势识别跟踪可以分为手部的定位和手指的跟踪。它涉及到对颜色空间的选取，以及由点到面再回到点的分析方法。首先，从含有手的场景中检测出手，并且得到手所在的大致区域，提取出手掌的中心位置作为追踪的前提。然后，应用模糊点脊模型来描述手，对于基本的伸展状态进行了分类，为语义识别打下基础。为了控制不确定环境条件的影响，提出了一种直方图匹配算法以及使用搜索窗搜索定位方法。整个试验分成手指定位和手部区域跟踪，得到的最终数据是手指的伸展方向和角度。在有限的限定条件下，系统的初始化算法能得到更多的手部特征，相对于基于神经的方法减少了训练强度，但又保持了其准确智能的特点，能够满足实时的要求。实验的目的在于尽量减少人工干预，结果证明：这种方法的可靠性和效率都有一定提高，具有研究价值。

英文摘要：

Compared with other systems in need of third-part equipment, the vision-based gesture tracking system is more convenient and natural. Single camera based gesture tracking includes locating of the hand and tracking of the fingertip. The algorithm is concerned with selection of color spaces and analysis from point to plane and back from plane to point. Firstly, the hand and the rough area of its location are checked out from the scene, and the center of the palm is extracted to facilitate further tracking. Then, a BR model is built to express the postures of the hand and the usual extending states are classified to lay foundation for sematic recognition. A new histogram matching and window searching ＆ locating tracking algorithm is proposed in real-time application to avoid the influence of negative conditions from the scene. The experiment consists of fingertip locating and hand area tracking. The final data results demonstrate the directions and angles of fingertip at different time. Compared with the neural method, the experiment gets more hand features, maintains intelligent accuracy and reduces the training strength under finite limited conditions. For a goal of less manual intervening, it achieves real-time processing. The results prove the algorithm has high reliability and efficiency and is worth researching.