Research On Target Tracking And Retrieval Technology In Surveillance Video

At present,intelligent video surveillance is widely used in public security monitoring,residential area monitoring,industrial site monitoring,traffic condition monitoring and other monitoring of the scene,constituting the core of city public security monitoring platform.In this thesis,we focus on the research of target tracking and search technology in video surveillance,which has important significance and broad application prospects.The thesis first introduces the research background and current situation of intelligent video surveillance,then focuses on the basic theory of target tracking under single camera,presents a robust tracking algorithm,describes the content-based image retrieval theory,and finally gives the image retrieval system.The main work of this thesis is as follows:(1)Due to the complex background,the change of illumination and the relation between objects make multi-target tracking robustness poor.To solve the problems,In this thesis,it is put forward a globally multiple maximum clique graphs based on the hierarchical multi-target tracking algorithm.It is based on the motion model and appearance model to create diagram,dealing with blocking targets with the help of aggregated dummy nodes,adding space constraints to solve the problem of identity conversion.Mixed-binary-integer programming is applyed to work out the problem on the overlapping fragments,at the same time getting more great group.The experiments are conducted on public datasets and the various datasets are increased.(2)Image retrieval system is designed on the basis of multiple features fusion.First the image color,texture and shape features are extracted using Euclidean distance similarity measure of the fusion feature vector after its external and internal normalization.Secondly,The design of the interface of the retrieval system is simple and friendly,users load the video at first,after target detection and image preprocessing,the image is stored in the image database,and its feature vector are stored in the feature database by feature extraction.When selecting a image from the database,a series of images are supplied by matching technology,choosing the first six pictures of the highest matching degree as the final output.The experimental results show that the system has a high recall ratio and precision ratio,which has great practical significance.