[1]杨岳航,孙涵.基于部件模型的复杂场景车辆检测方法[J].计算机技术与发展,2018,28(12):34-37.[doi:10．3969/j． issn．1673－629X．2018．12．007]
　YANG Yuehang,SUN Han.A Vehicle Detection Method Based on Part Model in Complex Traffic Scenes[J].,2018,28(12):34-37.[doi:10．3969/j． issn．1673－629X．2018．12．007]
点击复制

基于部件模型的复杂场景车辆检测方法()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
28
期数:
2018年12期
页码:
34-37
栏目:
智能、算法、系统工程
出版日期:
2018-12-10

文章信息/Info

Title:
A Vehicle Detection Method Based on Part Model in Complex Traffic Scenes
文章编号:
1673－629X(2018)12－0034－04
作者:
杨岳航;孙涵;
南京航空航天大学 计算机科学与技术学院，江苏 南京 211106
Author(s):
YANG Yue－hang; SUN Han
School of Computer Science and Technology，Nanjing University of Aeronautics and Astronautics，Nanjing 211106，China
关键词:
部件模型; AND-OR模板; 车辆遮挡; 车辆检测; 特征提取; 训练模型
Keywords:
part－based model; AND－OR template; vehicle occlusion; vehicle detection; feature extraction; model training
分类号:
TP301.6
DOI:
10．3969/j． issn．1673－629X．2018．12．007
摘要:
复杂场景中车辆间的遮挡,会造成车辆视觉信息损失,致使车辆出现漏检问题。为解决此问题,提出了一种基于部件模型的复杂场景车辆检测方法。首先,根据车辆图像的易遮挡程度将车辆对象分为易遮挡区域和通常可见区域两个部分,用于构建车辆部件模型;其次,通过运用Gabor滤波器对训练图像滤波,得到部件特征图;使用AND-OR模板分别对易遮挡区域与通常可见区域建模,根据训练特征图生成对应部件模型,以及部件模型之间的位置和尺度关系;最后根据部件模型检测图像中的候选部件,并根据部件间的位置和尺度关系筛选组合生成车辆目标描述,实现车辆检测。经实验验证,该方法漏检率低,并且能够有效应对遮挡与车辆姿态的变化。
Abstract:
The occlusion between vehicles in complex traffic scenes will cause the loss of visual information of vehicles，resulting in the problem of missing detection． In order to solve the problem，we propose a vehicle detection method based on part model． Firstly，a vehi- cle is divided into two parts，an easily－occluded region and a commonly－visible region which are used to construct part－based model． Secondly，the training images are filtered by Gabor filter to obtain the model graph． Then the two parts are modeled by the AND－OR template． At the same time，the probability distribution of the position and scale between the two parts models is calculated． Based on the training model，the candidates of such parts are detected，and the position and scale relation between the component parts are used to refine and generate the vehicle representation． The experiment shows that the proposed method effectively deals with vehicle occlusion and de- formation with low detection precision．
更新日期/Last Update: 2018-12-10