基于多信息融合的车辆阴影检测与车辆跟踪算法  

Vehicle shadow detection and vehicle tracking algorithm based on multiple information fusion

作　　者：曹晓娟[1,2] 王文涛[1] 宋晓琳[1] 张伟伟[1] 

机构地区：[1]湖南大学汽车车身先进设计制造国家重点实验室,湖南长沙410082 [2]湖南机电职业技术学院汽车工程学院,湖南长沙410151

出　　处：《中南大学学报（自然科学版）》2015年第11期4049-4055,共7页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(51175159);中国博士后科学基金资助项目(20110490263);汽车车身先进设计制造国家重点实验室自主课题团队重点资助项目(61075004);湖南省教育厅科学研究资助项目(13C261)~~

摘　　要：为了准确检测车辆,提出一种基于颜色、纹理、光照模型相结合的阴影检测算法。根据颜色恒常性完成阴影的初步检测,利用局部二值模式(LBP)纹理不变性和基于光照模型的亮度比值置信区间去除误检阴影像素,最后用区域生长完成阴影边缘像素的恢复,保证车辆阴影检测的准确与完整性。为了保证不同智能监控场景下车辆追踪的准确度和稳定度,提出一种特征与概率相结合的改进的Camshift跟踪算法。研究结果表明:所提出的阴影检测算法与改进的Camshift算法可以提升车辆检测与跟踪的准确性与稳定性。To make vehicle detection accurate, a shadow detection algorithm which combined color, texture and illumination model was proposed. Initial shadow detection was completed by the color invariance. LBP （local binary pattern） texture invariance and the brightness of the lighting model confidence interval were used to get rid of false shadow pixels. Finally, the recovery of edge pixels was realized by using region growing. And shadow detection was made more accurate and integrated by the algorithm. To make the vehicle tracking under different intelligent monitoring scenes more accurate and stable, improved Camshift tracking algorithm which combines characteristic and probabilistic was used. The results show that the proposed algorithm can improve the accuracy and stability of vehicle detection and tracking.

关 键 词：阴影检测 LBP纹理 车辆空间运动概率 CAMSHIFT算法 车辆追踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]