为研究目标跟踪问题，构造多种不同尺度的滤波器对目标进行滤波，生成目标高维特征；利用符合有限等距性质要求的稀疏矩阵对高维特征进行采样，获得目标低维特征。采用朴素贝叶斯分类器输出与Bhattacharyya 系数乘积的形式作为目标与候选目标之间的相似性度量，并选择最大值所对应的候选目标作为下一帧中的目标。文中提出了一种改进的快速压缩跟踪算法。实验表明，该改进的算法能够对目标进行有效跟踪。%An improved fast compressive tracking algorithm is proposed in view of the issue of object tracking . Firstly, multiple filters with various scales are constructed , which are used to produce target′s high-dimensional feature by filtering .Secondly , a sparse matrix that satisfies restricted isometry property is employed to generate low-dimensional feature of the target by sampling the high-dimensional one .Finally, product of a naive Bayes classifier′s output and Bhattacharyya coefficient is adopted to measure the similarity between the target and its candidates , the candidate with the maximum product value being selected as the target in the next frame .The experiments show that the proposed method could track the target more efficiently .
展开▼
