摘要
随着交通视频监控系统的广泛应用,车辆检索越来越受到公共安全管理部门的重视。车辆检索自动化系统是以图搜图研究的重要应用方面,成为替代人工检索的重要工具。该系统可以将人从繁杂的检索工作中解放出来。文章提出一种多特征融合的车辆检索匹配方法,这些特征包括色度直方图、饱和度直方图、明度直方图、灰度直方图,感知哈希汉明距离,颜色布局描述符以及SIFT(尺度不变特征变换)特征。实验结果表明:与仅使用颜色和仅使用纹理的方法相比,通过融合多种颜色表示方法和形状纹理特征,使检索结果更加准确,进一步减少了人工从检索结果中筛选需要的信息的工作量。
With the wide application of video traffic monitoring systems,the vehicle retrieval is getting more and more attention from the public safety administration.Automated vehicle retrieval system is an important application of the content-based image retrieval,and it becomes an important tool to replace the manual ones.Human power can be liberated from the complex search work to do more important jobs.We propose a vehicle retrieval method based on multi-feature fusion,including histograms of hue,saturation,value and gray images,the distance of Perceptual Hash Hamming,Color Layout Descriptor and SIFT key point matching algorithms.The experimental results show that the results of proposed method fusing multi-color-representation and shape-texture features are more accurate,compared with the color-based method and texture-based method,and the proposed method further reduces manual screening important information workload from the search results.
作者
梁中岩
郭乔进
胡杰
Liang Zhongyan Guo Qiaojin Hu Jie(The 28th Research Institute of China Electronics Technology Group Corporation, Nanjing210007, China)
出处
《信息化研究》
2016年第4期51-58,共8页
INFORMATIZATION RESEARCH
