基于摄像机的人体运动分析系统标志点图像处理

Landmark image processing for human motion analysis system based on video camera

针对基于普通摄像技术的人体运动分析系统检测精度与处理速度低的问题，提出了一种快速标志点模糊图像预处理及标志点识别与跟踪的新方法。该方法首先利用快速模糊边缘检测算法对图像进行增强处理，然后使用作者构造的基于阈值法的连通域判别算法提取标志点图像边缘，最后运用“一阶预测—局部搜索”方法查找与跟踪标志点。对比实验及临床应用结果表明，该方法具有运算量小，图像处理快的特点，提高了运动分析系统的运行速度。该方法尽管是针对由普通摄像机替代专用设备建造的人体运动信息分析系统中图像处理问题提出的，但它同样适用于其它有关静止或低速运动物体图像的检测、处理与模式识别等方面的研究。

A new human motion analysis system based on common video camera was described. In this system a special landmark image processing method was introduced which largely reduced the cost of image computation. This method includes three steps: firstly, a fast algorithm for edge detection using fuzzy sets was employed to enhance the images; secondly, an area detecting algorithm was used to get the edge of landmarks; finally, one order predicting and local searching algorithm was adopted to find and track landmarks. Experimental results and clinical application were presented to demonstrate the advantages of this new landmark image processing method. Although this method is proposed for the human motion analysis system based on common video camera, it is available to other detection and analysis of low speed moving objects.