基于Hausdorff跟踪器的人脸跟踪识别

Face tracking and recognition based on Hausdorff-based tracker

Snake算法能够跟踪运动图像中对象的非刚性运动,但是对于背景复杂的图像,Snake跟踪的结果不够理想。因而在首帧分割得到对象轮廓的二值模型后,再采用基于Hausdorff距离的跟踪器,找到对象模型在后继帧中的最佳匹配位置;然后采用Snake模型对该匹配位置上的非刚性形变的像素进行匹配。实验表明:对于具有静止背景且前景对象不是快速运动的视频序列,与直接采用Snake技术进行运动对象的跟踪相比,该提取视频对象平面过程能够进一步提高结果的正确性。

Snake algorithm is capable of tracking moving object in image no rigid motion, but for images with complicated background, Snake tracking results are not ideal. Thus in the first frame segmentation object contour two value model, and then based on Hausdorr distance tracker, to find the object model in subsequent frames in the optimal position; then uses Snake model to the matching position of the non- rigid deformation pixel and foreground object matching. The experimental results show that： in view of the static background is not a fast motion in a video sequence, and directly adopts the Snake technology of moving object tracking in the extraction process, which can further improve the correctness of the results.