针对传统目标跟踪算法判别力及稳健性不足的问题，本文在对跟踪输出响应图可信度进行充分研究的基础上，结合目标尺度估计方法，提出多特征融合和自适应尺度估计相结合的目标跟踪算法。该方法通过计算不同特征模型下的输出响应图可信度，实现对两种互补的特征进行自适应加权融合，有效地提升了表观模型的鉴别力及泛化性能。尺度估计模块通过构建多分辨率特征金字塔、训练尺度滤波器及尺度特征降维，避免了在尺度空间内的穷举式搜索。实验表明文中算法有效地提升了跟踪过程中的准确率和成功率，能够适应遮挡、形变等复杂场景下的目标跟踪，并且具有非常高的效率。

To improve the discriminability and robustness of appearance models in traditional tracking algorithms, in this study, we analyze the confidence of output response maps and propose a tracking algorithm with adaptive weighted multiple feature fusion approach and scale estimation module. By integrating two complementary features, the robustness and generalization ability of the appearance model were significantly improved. The scale estimation module avoids exhaustive searching in scale space and effectively reduces the computational cost by constructing a multi-resolution feature pyramid, training scale filter and reducing the dimensionality of scale features. Experimental results show that the proposed tracking algorithm can effectively improve the tracking performance in terms of precision rate and success rate, and the tracker can perform with high efficiency when the object is under occlusion and deformation.