基于动态规划与填充密度的监控视频摘要算法  

Monitoring Video Abstract Algorithm Based on Dynamic Programming and Filling Density

作　　者：丛超 CONG Chao(School of Electrical and Electronic Engineering,Chongqing University of Technology,Chongqing 400054,Chin)

机构地区：[1]重庆理工大学电气与电子工程学院,重庆400054

出　　处：《计算机工程》2018年第7期250-258,共9页Computer Engineering

基　　金：重庆市自然科学基金(cstc2016jcyj A0436);重庆市教委科学技术研究项目(KJ1709210);重庆理工大学项目(2015XH13)

摘　　要：传统监控视频摘要算法存在空间效率较低、浓缩长度较为固定的问题。为此,提出一种基于活动对象捕捉的监控视频浓缩与视频摘要生成算法。定义填充密度的概念,并将其引入能量函数模型,用于定义视频浓缩的饱和程度。在能量函数最小化的过程中,利用动态规划方法设计活动片段组合优化策略,从而解决视频浓缩长度确定问题。实验结果表明,与基于活动管道的摘要算法相比,该算法在保证高浓缩比的前提下,减少了运动物体的丢失率,提高了运行速度,并具有较高的鲁棒性。In surveillance video synopsis field, some traditional algorithms have defects of low spatial efficiency and a stationary condensation length. Therefore, an innovative algorithm based on active object tracking is proposed. Firstly the concept of packing density is established and introduced into energy function model. Then,in order to solve the efficiency problem of energy function minimization,a dynamic programming method is proposed to determine the length of video compression and an optimized arranging strategy is created to reduce the computing cost of energy minimization.Experimental results show that, comparing with traditional Tube based algorithms,this algorithm not only reduces the loss rate of moving objects,but also improves the running speed with a remarkably high robustness and practicability.

关 键 词：监控视频 图像处理 视频摘要 视频浓缩 动态规划 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]