基于混合差分的车辆检测方法  

Vehicle detection algorithm based on hybrid difference

作　　者：朱娟[1] 陈杰[1] 

机构地区：[1]华南理工大学计算机科学与工程学院,广东广州510006

出　　处：《计算机工程与设计》2011年第1期332-335,共4页Computer Engineering and Design

摘　　要：以实时获取的视频图像为基础,针对ITS领域的关键技术,对高速公路事件检测系统的多个方法进行了改进。在彩色图像模型下提出一种改进的基于混合差分的车辆检测方法,使检测更快速,效果更好。针对检测结果的后期处理,提出了一种实时分割与实时合并移动目标的算法。实时分割主要采用一种基于阴影消除的方法,通过模板匹配,从而达到分割重叠车辆的目的。实时合并则引进一种兴趣度函数来填补移动目标的内部空洞,同时通过测定六边形之间的距离使断裂部分进行合并。实验结果表明,提出的算法简单、易实现、具有高实时性,车辆检测率超过96%。According to the key techniques of ITS filed,several improved techniques in free way incident detection system are studied and analyzed on the basis of real-time video images.A hybrid difference strategy based on difference model of colored images is introduced to attain higher detecting quality than other methods.As regard to post-treatment of the test results,an algorithm for real-time segmentation and integration of the moving target is presented.The real-time segmentation mainly adopts an improved algorithm based on the elimination of the shadow,to divide the moving target and by matching the template,separate the overlapping vehicles.In the time of integration,this essay brings forth a type of interest function to make a real-time filling of the internal hollowness of the moving target,and have a real-time merger by making use of the distances between the hexagon.Experimental results shown that the various proposed algorithm is characterized by simplicity,easy-to-use and high real-time.The vehicle detection rate is over 96%.

关 键 词：车辆检测 混合差分 实时分割 车辆跟踪 事件检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]