基于可变模型的目标实时跟踪

为了在图像序列中实现目标的快速定位和实时跟踪，该文提出了一种基于可变模型的快速目标跟踪算法，在已知模型条件下，利用区域模型相关匹配的思想对目标模型进行实时更新，充分利用目标莲续运动过程中目标形状在两个连续帧中变化不大、相邻两帧中目标的速度和位移变化不大的特点，以当前帧目标模型作为下一帧的先验模型；综合运用模型梯度信息、运动信息和模型区域特征匹配的方法来跟踪目标。由于算法综合考虑了目标模型的区域信息和轮廓信息，因此对背景干扰不太敏感。在头部跟踪实验过程中，该文算法跟踪移动目标的实时性和准确性比较好，抗干扰能力较强，基本上可以满足鲁棒性和快速性的要求。

Real-time Object Tracking Based on Deformable Template

In order to achieve fast object locating and real - time object tracking in image sequences, a method for rapidly tracking moving object based on deformable template is proposed. Under the circumstance of known template, we update the object's template using the thought of region template correlation. The motion and deformation between two successive frames are not significantly large so the template in the current frame can be used as a prior template for the next frame. We use the template's image gradient and motion cues combined with the region feature matching to track the object. Because our method concerns both the object region and object boundary cue , it is insensitive to background clutter. In the head tracking experiment, the method discussed in this paper performs well in fast and accurate tracking of moving object and also has a high quality of anti -jamming, which could basically meet the requirement of robustness and swiftness.