基于DWT-DCT系数符号特征匹配的视频目标跟踪

Moving Object Tracking Via DWT-DCT Coefficient Symbol Feature Matching

视频运动目标跟踪,简单说就是在下一帧图像中锁定感兴趣目标的确切位置。复杂的背景以及目标本身的变化给运动目标跟踪技术带来了很大的困难。现有算法大都在分辨率较高的条件下针对特定场景取得较好的跟踪效果,但针对运动目标尺寸比较小且分辨率较低的目标跟踪算法的研究报道不多,这种情况下通常难以达到精确的跟踪,鲁棒性也比较差。针对这一问题,探讨了一种基于DWT-DCT系数符号特征的运动目标跟踪算法,该算法通过提取这一特征进行特征匹配来定位目标的位置。实验结果表明,该算法不仅具有较强的鲁棒性,而且在低分辨条件下仍能对尺度较小的目标进行精确的跟踪。

Moving object tracking in video is to locate the exact position of an interested object in the next frame image. The complex background and uncertainty of object itself will bring great challenge to object tracking technology. Most of the existing algorithms,which obtained satisfactory results in certain typical application conditions,are based on the conditions of the relatively high resolution on a typical application scene. However,research work for small size object tracking in low-resolution video is relatively less,and it is difficult to achieve accurate tracking if the general algorithms are applied to this situation. Therefore,research of efficient algorithms for object tracking in low resolution video is a more challenging work. In this manuscript,an algorithm of this type was proposed by using DWT-DCT coefficient symbolic features as matching feature. Experiments were provided to demonstrate the feasibility of the proposed algorithm for object tracking in low resolution video.