多监控摄像头特定目标检索方法研究

摘要

Video surveillance is an important part of security system. With the upgrade of security standard and requirements, more and more technologies areused in the video surveillance system, for example, the pattern recognition technology based on video image processing is widely used in video surveillance, and this of more important practical significance for the searching of a specific target under the multiple surveillance cameras. The existing methods for object retrieval, are analyzed, and in combination with the DPM (Deformable Part Model), the new retrieval method suitable for specific object under multiple surveillance cameras is also proposed. By extracting multiple object sequences from the different surveillance videos with the involved object detection and object tracking algorithms, and comparing these object sequences with the retrieval object, the retrieval information of the retrieval object in different surveillance videos could be acquired.%视频监控是安全防范系统的重要组成部分.随着安防指标和需求的提高,越来越多的技术被应用到视频监控系统中,如基于视频图像处理的模式识别技术已广泛应用于视频监控中.而在多监控摄像头环境下,对特定目标的检索往往具有更加重要的意义.因此,分析现有目标检索方法,结合Deformable Part Model(DPM)目标检测模型,提出一种新的适用于多监控摄像头目标检索的方法.该方法包含目标检测和目标跟踪两个模块,通过提取不同监控录像中的多个目标序列,并将检索目标与之比对,获得该检索目标在不同监控视频中的检索信息.