摘要
前景物体的边缘信息对混合交通视频检测的参数提取和表达具有非常重要的作用.为克服孤立地利用当前图像无法准确提取边缘信息的问题,提出了基于时空上下文表达的混合交通边缘提取新算法.在获取当前边缘信息基础上,结合运动目标的特征属性与时空上下文相关信息,通过检索文本的关联性,进行前景边缘提取.实验表明,文中算法能够准确而充分利用上下文信息实现前景边缘的提取,前景边缘的有效提取率可达95%以上,为后续混合交通视频检测的分类识别和参数提取提供了有效的依据.
Foreground edge information has a very important role in the parameter extraction and the expression of mixed traffic video detection. To overcome the problem that the accurate edge information cannot be extracted from current frame solely, a new edge extraction algorithm for mixed traffic based on spatial and temporal context is proposed. Based on the current frame edge information, combining the characteristics of the moving objects with the information of spatial and temporal context and retrieving the relevance of the text, the foreground edge is extracted. Experimental results show that the algorithm can get the extraction of the edge of the prospects accurately and adequately using context, and the effective extraction rate can reach above 95G, which provide an efficient basis for the subsequent classification and parameter extraction of the mixed traffic video detection.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2013年第12期1289-1292,1297,共5页
Transactions of Beijing Institute of Technology
基金
国家"八六三"计划项目(2011AA110304)
关键词
混合交通
视频检测
时空上下文
mixed traffic
video detection
spatial and temporal context
