摘要
稀疏表示已经成为运动目标检测的有效方法之一,但其还没有很好地解决目标检测的快速性和鲁棒性.本文基于最大后验概率提出了一种快速鲁棒的运动目标检测模型,并设计了该模型的求解算法.该算法包括两个阶段:在第一阶段利用编码迁移实现稀疏系数的快速求解;在第二阶段基于运动目标的空间连续性结构,利用图切实现目标检测.在多个具有挑战性的图像序列上的实验结果表明,与其他经典运动目标检测算法相比,本文方法在快速性和鲁棒性方面具有较优的性能.
Sparse representation is one of effective methods in dealing with the moving object detection. However,the quickness and robustness of object detection are far from being solved in the existing methods. In this paper,a fast and robust moving object detection model based on the maximum posteriori probability is proposed,and a two-stage detection algorithms is designed. At the first stage,sparse coefficient is quickly solved by using coding transfer; At the second stage,based on spatial continuity structure,moving object detection is achieved by using graph cut. The experimental results on several challenging image sequences showthat the proposed method has better performance than the existing classical moving object detection algorithms in rapidity and robustness.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2017年第10期2355-2361,共7页
Acta Electronica Sinica
基金
安徽省自然科学基金(No.1508085QF114
No.1608085QF144)
国家自然科学基金(No.61379105)
中国博士后科学基金(No.2014M562535)
关键词
运动目标检测
稀疏表示
编码迁移
图切
moving object detection
sparse representation
coding transfer
graph cut
