一种面向高斯差分图的压缩感知目标跟踪算法

传统的压缩感知目标跟踪在目标纹理改变、比例缩放、光照变化剧烈时鲁棒性不足，本文提出一种面向高斯差分图的压缩感知目标跟踪算法。首先，构建原始图像的多尺度空间及其对应的图像高斯差分图，实现高斯差分图的特征提取并获取压缩感知的输入信号；然后，通过压缩降维，目标邻域遍历，参数更新等过程，计算出面向高斯差分图的后续帧的目标最优跟踪窗；最后，将跟踪窗投影到对应的原始图像上，完成面向视频流的目标跟踪。实验证明，高斯差分图像是单通道灰度图，相比较原始视频流的三通道彩色图，具有灰度取值范围小，数值低，结构简单，维数少等特点，增强了特征对纹理改变、比例缩放和光照变化的稳健性，且继承了原始算法的实时性。因此，与传统的压缩感知算法相比，本文算法能快速准确地实现复杂环境下的移动目标跟踪任务，具有更强的鲁棒性。

Target tracking by compressive sensing based on Gaussian differential graph

The traditional target tracking algorithm of compressive sensing has poor robustness in texture change, scale variation and illumination change. In order to solve it, the paper proposed a novel target tracking algorithm using by impressive sensing based on Gaussian differential graph. Firstly, multi-scale space of image is constructed as to acquire Gaussian differential graph. The features are extracted on the Gaussian differential graph and are taken as input signals of impressive sensing. Secondly, by compressing, dimension reduction, target neighborhood traversal, parameters update, the optimal search-window of sequential Gaussian differences graph is estimated. Thirdly, the search-window is mapped onto the corresponding original image, and target tracking is finished effectively in the video sequences. Compared with the three-channel original image, experiments proved that, the Gaussian differential graph had some characteristics such as single-channel, small grayscale range, low value, simple structure, small dimensions. The robustness was enhanced in the scaling, texture and illumination changing, and the real-time performance was inherited from the traditional algorithm. Finally, compared with traditional compressive sensing, the proposed algorithm was used well in real-time moving target tracking and had better robustness in complex environment.