基于压缩感知的多特征加权目标跟踪算法

针对被跟踪目标运动、纹理或环境变化时, 采用基于压缩感知目标跟踪算法目标易漂移、丢失的问题, 提出了改进的压缩感知目标跟踪算法。通过压缩感知算法提取灰度和纹理特征, 计算特征对样本分类结果并更新特征的权值, 使用加权过的特征寻找目标在下一帧的位置。对不同视频的测试结果表明, 提出的算法在目标运动、纹理或环境变化的情况下跟踪准确, 在目标大小80×120像素时平均帧速为25 fps。与传统的压缩感知跟踪算法和其他跟踪算法相比, 所提出的算法在目标运动、纹理或环境变化时能快速准确地获取跟踪目标, 并具有更强的鲁棒性。

Real-time tracking using multi-feature weighting based on compressive sensing

As traditional tracking algorithm based on compressive sensing is failed to track target stably when the target quickly move, texture or environment get seriously changed, this paper proposed a reformative target tracking algorithm based on compressive sensing. By extracting the gray and texture features using compressed sensing, it calculated the feature's weight according to the classification results, then found the target in next frame. Results of tests on variant video sequences show that the proposed algorithm is capable of speedily and accurately capturing the tracking target for target moved, texture or environment changed, and average computing frame rate is 25 fps when the target scale is 80 ×120 pixel. As compared with the tracking algorithm based on compressive sensing and other tracking algorithms, the proposed algorithm can hold a better robustness in target moved, texture and environment changed.