Bayonet Spatially And Temporally Correlated Adjacent Vehicle Video Retrieval

In order to escape punishment, drivers always take methods such as sheltering, staining, forging and changing motor vehicle plates as well as not hanging the license plate to let the current recognition system for the vehicle license plate not realize the effective identification on the license plate. Meanwhile, every day, there are quite large amount of such traffic law violation vehicles that cannot be conducted the statistics, analyzed and handled manually. For the causes above, it brings about huge difficulty for the traffic enforcement. How to make the associative lookup of the vehicle objects from the surveillance video has been one of hot issues to be solved for traffic management law-enforcing department.The paper deeply studies the spatiotemporal association retrieval of the video images of vehicle objects in the traffic monitoring. Firstly, it makes the research on the image preprocessing, laying the foundation for the feature extraction. Secondly, the feature selection and extraction are done. Its emphasis is to investigate characteristics such as color, geometry and outline, providing services for the similarity matching analysis of the vehicle object images. Next, the similarity matching analysis of vehicle objects is introduced. It first makes the classification on vehicles and then demonstrates the similarity matching analysis of vehicles based on the multi-feature fusion. Finally, according to the similarity result, the feedback of the relevant video is delivered to users. The main research tasks of this paper consist of several points below:First, introduce the image preprocessing technology and algorithm. Through the research on extraction of key frame of video, image normalization, image smoothing processing, image graying&equalization processing as well as operations and experimental analyses including the vehicle detection&segmentation and edge detection, the image preprocessing technology and algorithm fit for the research of this paper are achieved, which lays the solid foundation for the subsequent extraction of vehicle image features. Secondly, feature selection and extraction. Aiming to the features of vehicles in the traffic surveillance videos and images snapshot, it selects the vehicles’features, focusing on interpreting the extraction methods of features such as the color of vehicles, geometry and outline.Thirdly, the similarity matching study of the spatiotemporal association vehicle objects. First of all, the speed of the vehicle objects is researched. It proposes the speed detection method based on the video images. In the similarity matching study, it firstly classifies the vehicles, putting forward the similarity matching method on the basis of multi-feature fusion. It conducts the similarity matching research by integrating various features such as the color and geometry.Fourthly, the research of the video retrieval. Based on the study on the image preprocessing, feature extraction and similarity matching, the video retrieval of vehicles of spatiotemporal association is discussed. It designs the video retrieval procedure and system.Fifthly, the relevant algorithm and methods included in this paper. The corresponding experiment is presented. Besides, it summarizes and analyzes the result of the experiment.