基于高斯背景模型的车辆检测改进算法  

Improved vehicle detection algorithm based on Gaussian background model

作　　者：胡学刚[1] 刘忠振[2] 

机构地区：[1]重庆邮电大学数理学院,重庆400065 [2]重庆邮电大学计算机学院,重庆400065

出　　处：《计算机工程与设计》2011年第12期4111-4114,共4页Computer Engineering and Design

基　　金：国家自然科学基金项目(11071266);重庆市教委科研基金项目(KJ100505);重庆邮电大学自然科学基金项目(A2009-31)

摘　　要：针对车辆检测中实时性、准确性以及自适应性很难兼顾的问题,提出了一种快速有效的车辆检测算法。该算法采用阈值判定背景区域和更新区域,根据背景变化的程度,利用动态权值更新学习率,使用基于单高斯背景模型方法,结合逻辑"或"运算检测出车辆。实验结果表明,该算法能够快速准确地检测出车辆,并且在背景更新过程中实现了自适应,并具有一定的鲁棒性。For the problem that it is difficult to take into account real-time,accuracy and adaptability together in vehicle detection,a fast and effective vehicle detection algorithm is proposed.By using threshold values,the background area and the update region are determined,and by using dynamic weights,learning rates are updated in accordance with the degree of changes in background.With logical ＂ or ＂ operator,vehicles are detected based on a single Gaussian background model.The experimental results show that this algorithm can detect vehicles quickly and accurately,achieve adaptive background updating and have certain robustness.

关 键 词：车辆检测 单高斯背景模型 背景更新 背景差分 二次帧差 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]