基于视频多特征融合的列车车厢烟雾检测方法  

Fire smoke detection in train carriages based on multiple features fusion

作　　者：陈海鹏[1] 刘飞[1] 徐磊[1] 徐广伟[1] 王宏雷 谭华春[3] 

机构地区：[1]南车青岛四方机车车辆股份有限公司,山东青岛266111 [2]易程(苏州)智能系统有限公司,江苏苏州215163 [3]北京理工大学机械与车辆学院,北京100081

出　　处：《北京交通大学学报》2015年第1期67-71,77,共6页JOURNAL OF BEIJING JIAOTONG UNIVERSITY

基　　金：国家"863"计划项目资助(2012AA112001)

摘　　要：列车车厢中环境相对复杂且遮挡物较多,为了提高视频烟雾检测的准确性,剔除场景中疑似烟雾的干扰物,降低由于环境中光照变化导致的误检情况,提出了基于运动、颜色及衰减等多特征融合的烟雾检测算法.该算法能有效克服背景复杂、光照变化大,以及阴影问题带来的干扰.算法主要由3部分构成:烟雾运动检测、颜色特征分析提取和多特征融合.该算法通过背景建模方法分割出运动的区域,引入归一化RGB空间烟雾颜色基础模型及衰减模型,以剔除疑似烟雾区域的干扰,同时可以有效地降低光照影响.Owing to complex environment of the train carriage,in order to improve the accuracy of the video based on smoke detection algorithms,eliminate the similar object of smoke movement and reduce the false detection rate caused by environment light changes,a modified smoke detection algorithm based on motion,color and attenuation is presented. This algorithm can effectively reduce the interference of background motion and the changing scene illumination,and eliminate the shadow phenomenon. The algorithm consists of three parts： the detection of smoke movement,extracting the characteristics of smoke color and multiple features fusion. First,the movement of the pixels is detected by the motion segmentation. Then the model of normalized RGB color space is introduced for eliminating the interference of similar smoke area and reducing the impacts of changes in illumination effectively.

关 键 词：图像处理 视频烟雾检测 运动分割 衰减模型 多特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]