中文摘要：

ASM是一种应用于非刚体轮廓匹配的统计模型方法,由于匹配结果的可靠性依赖于先验的灰度模型,导致灰度信息发生动态变化时基于ASM的目标跟踪效果不佳.针对跟踪中轮廓匹配的鲁棒性问题,提出了一种基于改进ASM的目标跟踪方法,该方法采用一种在线提取和更新灰度模型的机制,摆脱对目标先验灰度的依赖;并结合强边缘特征和目标内部灰度信息,改进传统ASM方法的灰度模型和搜索算法,提高了运动过程中目标附近背景信息变化情况下轮廓匹配的鲁棒性和快速性;跟踪过程中利用卡尔曼滤波预测目标位置提高了运行效率.实验验证了方法的有效性和鲁棒性.

英文摘要：

ASM is a model based method which is applied to matching contours of non-rigid objects. Due to their dependence on the prior grey level model, the tracking results may be ineffective when grey level information changes dynamically. To improve the robustness of contour matching when tracking moving objects, a novel method based on ASM was proposed. The method got rid of the dependence on the prior information by learning the grey level model online. Meanwhile, in order to improve the accuracy, target inner grey information was adopted to replace the traditional model and the strong edge feature was used to enhance the performance of the iterative search algorithm. The proposed method also combined Kalman filter to improve the efficiency. Experiments show its validity and robustness.