改进的KSMS手势跟踪算法研究  

An improved KSMS hand tracking algorithm

作　　者：丁辉[1] 纪成双[1] 王琳[1] 

机构地区：[1]首都师范大学信息工程学院,北京100084

出　　处：《光学技术》2013年第5期456-459,共4页Optical Technique

基　　金：国家自然科学基金项目(11178017);北京市自然科学基金资助项目(4132014);北京市教委科技面上项目(KM201110028017);首都师范大学实验室开放基金项目

摘　　要：Mean-Shift算法是一种基于颜色信息的有效跟踪算法,当背景中存在相似颜色信息干扰时会导致目标跟踪丢失。针对手势识别中的肤色干扰问题,提出了一种改进的基于肤色约束的Kalman滤波和Mean-Shift融合KSMS(Kalman Skin Mean-Shift)跟踪算法。利用Kalman滤波的预测功能,对搜索窗口提前进行预测,从而提高跟踪的准确性。实验结果表明,算法准确可靠,在大面积肤色信息干扰时具有良好的鲁棒性。The Mean-Shift algorithm is a commonly used method in the field of target tracking, but it will lost the ob ject when tracking target in complex background of similar color interference. For skin color interference in gesture track ing and recognition, based on skin color constraint, an improved Kalman filter and Mean-Shift fusion KSMS （Kalman Skin Mean-Shift） tracking algorithm is proposed. Using Kalman filtering prediction function to predict the window of searching ahead of time, the aceuracy of tracking is increased. Experimental results show that algorithm is accurate, relia ble and good robustness to deal with the skin color interference problems.

关 键 词：目标跟踪 KALMAN滤波 MEAN-SHIFT算法 肤色信息 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]