基于肤色CIELab空间的视觉窗口手势跟踪

CIELab Color Based Vision Window Hand Tracking

考虑到手势的应用场合,手部肤色作为特殊的识别标记往往区别于环境,因此加入肤色分割对手部图像序列进行分析以及跟踪.在CIELab空间,采用了平均转移分割算法提高了分割速度.在此基础上结合单目视觉中对指尖的识别,提高了手部分割的可识性和未来手部语义应用的可行性.跟踪则采用了卡尔曼滤波处理后的窗口跟踪.手指尖的半封闭特征的利用以及窗口算法保证了处理过程中的鲁棒性和准确性.

Considering the gesture's application occasion, the hand skin color is the conspicuous mark compared with environment around.The segment based on color to the hand image sequences is compiled to analyze and track. In CIELab color space, we use mean transition segment algorithm to improve the computational speed. For further recognition of sign and the reasonable hand segment,the analysis and tracking of fingertip is importable. The tracking algorithm is similar to Kalman filter with windows search. Actually, the occlusion of fingertip and window search reduced most computational cost. With the result, we can find the tracking is robust and quite satisfying.