基于冗余字典的多特征压缩感知目标跟踪算法

发布时间：2018-01-31 01:35

本文关键词： 信息处理技术冗余字典压缩感知粒子滤波目标跟踪　出处：《兵工学报》2017年06期 　论文类型：期刊论文

【摘要】：针对多特征压缩感知算法中,要求信号稀疏表示的基是一个正交矩阵的问题,提出了提取红外与可见光的多特征目标构造冗余字典子空间下的稀疏表示,分析了压缩感知算法中感知矩阵的选择和稀疏信号的重构。根据对信号稀疏表示的重构,提出粒子滤波框架下基于冗余字典的多特征压缩感知跟踪方法,能够自动检测复杂场景中出现的动态目标。实验结果表明,与其他经典算法相比,该算法在光照变化、相似外形的干扰目标遮挡等复杂场景中具有更好的鲁棒性及实时性。
[Abstract]:In order to solve the problem that the basis of signal sparse representation is an orthogonal matrix in multi-feature compression sensing algorithm, a sparse representation in redundant dictionary subspace based on extracting infrared and visible light is proposed. The selection of perceptual matrix and the reconstruction of sparse signal in compressed sensing algorithm are analyzed. According to the reconstruction of sparse representation of signals, a multi-feature compression perceptual tracking method based on redundant dictionaries is proposed in the framework of particle filter. It can automatically detect dynamic targets in complex scene. Experimental results show that compared with other classical algorithms, the algorithm changes in the illumination. It has better robustness and real-time performance in complex scenes such as object occlusion with similar shapes.
【作者单位】：南京理工大学自动化学院;南京理工大学紫金学院电子信息与光电技术学院;
【基金】：国家自然科学基金项目(U1330133)
【分类号】：TP391.41
【正文快照】： 0引言光电跟踪系统作为一种新型无源跟踪装备,其特点是采用光电传感器探测和跟踪目标,在复杂背景下迅速、可靠地提取目标影像,配合激光测距设备建立目标运动航迹,实现对目标连续稳定的状态估计和滤波,为各类火控系统实施火力打击提供精确解算的射击诸元信息[1]。因此对目标跟

【参考文献】