融合实时交通状态信息的道路背景更新模型  

A Road Background Updating Model Considering Real-time Traffic State Information Fusion

作　　者：佘永业[1,2,3] 李熙莹[1,2,3] 赵力萱[4] 余志[1,2,3] 

机构地区：[1]中山大学工学院,广东广州510275 [2]广东省智能交通系统重点实验室,广东广州510275 [3]视频图像智能分析与应用技术公安部重点实验室,广东广州510006 [4]广东警官学院,广东广州510230

出　　处：《公路交通科技》2014年第6期115-120,共6页Journal of Highway and Transportation Research and Development

基　　金：国家自然科学基金项目(51078362);公安部应用创新计划项目(2011YYCXGDST078)

摘　　要：针对道路背景的提取和更新效果受实时交通流状态影响的问题,提出了一种用于交通视频监控场景下的融合实时交通状态信息的道路背景更新模型和实现方法。该方法通过构建宏观交通状态参数特征集估计道路实时的交通状态信息,在传统方法基础上融合交通状态信息建立选择性背景帧提取模型,改进背景更新候选帧的选取条件,通过光照变化来估测道路背景的更新频率并自适应地提取背景更新区域。试验表明:该方法在道路交通顺畅和拥堵状态时,均能有效地提取道路背景,大大地提高背景更新的收敛速度。To cope with the problem that road background extraction and updating are affected by real-time traffic flow state, a road background updating model and implementing method considering real-time traffic state information fusion for traffic surveillance videos are proposed. The method estimates real-time road traffic state information by constructing macro traffic state parameters characteristic set. It establishes a selective background frame extraction model by fusing traffic state information on the basis of traditional methods. It improves the condition for selecting candidate frames of background updating, and estimates road background updating frequency by change of illumination and extracts background update region self- adaptively. The experimental result shows that road background can be effectively extracted both in the smooth and congestion states and the convergence rate of background updating can be greatly improved using the proposed method.

关 键 词：智能交通系统 背景更新 背景帧提取模型 交通状态检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]