摘要
压缩感知是信号处理领域的新理论,用于目标跟踪算法时可在大量底层特征中提取出少量重要信息,减少计算量,提高算法速度。传统的基于压缩感知的跟踪算法,为了保证算法速度,对压缩后的特征简单建模,准确性还有待提高。提出一种基于协方差矩阵的压缩感知跟踪算法,先利用压缩感知原理获取压缩后的Haar特征,再利用协方差矩阵融合Haar特征区域内的底层多维特征,以此构建目标模型,并通过搜索当前目标区域的邻域,利用流形空间上的距离度量算法匹配最佳目标,从而提高算法准确性。
Compressed sensing is a new theory in the signal processing.It can extract the important features from lots of low-level features when it is used for object tracking.Traditional tracking algorithm based on compressed sensing used a simple target model to keep the speed.To improve its accuracy,a compressed sensing tracking algorithm based on covariance matrix has been proposed.First,Haar features are compressed.by compressed sensing.Second,based on covariance matrix,a new target model with more low-level features is obtained.Then,the neighborhood of the current target is searched and the best target is matched using the manifold distance measure.Finally,the proposed algorithm gets a better accuracy.
出处
《软件导刊》
2017年第4期31-34,F0003,共5页
Software Guide
基金
国家自然科学基金项目(61170093)
湖北省教育厅科学技术研究计划重点项目(D20141603)
