针对压缩跟踪算法中存在的特征单一、对光照变化敏感、不能适应目标尺度变化的问题，提出了一种融合HOG类特征的尺度自适应压缩跟踪算法。该算法在Haar-like特征基础上采用固定比值方式融合了HOG类特征，降低了算法对光照的敏感度，HOG类特征仍采用积分图算法进行加速计算。另外，在尺度估计方面，采用相关滤波尺度估计方法，找到使尺度滤波器响应最大的尺度作为新的目标尺度，及时调整跟踪窗的大小并更新特征提取模板，解决了尺度变化问题。实验结果表明：改进后的算法可以适应较大光照变化与尺度变化、平面内旋转等情况，跟踪精度和鲁棒性有明显提高，并且满足准实时性的要求。

There are problems in traditional compressive tracking algorithms of few extracted features, high sensitivity to illumination changes and inadaptability to the changes in target scale. To solve these problems, an adaptive-scale compressive tracking algorithm fusing HOG-like features is proposed.On the basis of Haar-like features, the new algorithm uses the method of fixed ratio to fuse HOG-like features, so as to reduce its sensitivity to illumination. The integral graph algorithm is used to accelerate the calculation of HOG-like features.In addition, a scale estimation method based on correlation filter is proposed, so as to find out the scale to which the scale filter has the biggest response and then take it as the new target scale.The size of the tracking window is timely adjusted and the feature extraction template is updated.Therefore, the problem of scale change is solved. The experimental results show that the improved algorithm can adapt to marked illumination changes, scale changes and in-plane rotation.Its tracking accuracy and robustness are improved greatly, which satisfies the requirements of real-time tracking.