Motor Vehicles In Traffic Surveillance Video Retrieval Key Technology Research

At present, with the rapid development of our economy and the constant enlargement of the city, transportation has become the economic development of supply lines and appear increasingly busy, the number of motor vehicles is also in rapid growth. However, because of the traffic construction and the traffic law consciousness is relatively backward, traffic accidents happen frequently, this gives traffic supervision department has put forward higher requirements, motor vehicle retrieval technology widely used, not only for city traffic planning to provide first-hand data, also for traffic management department and the public security department to trace the vehicle to provide a great convenience.This paper gives a detailed analysis of content based video retrieval process, and studied the related algorithms and implementation technology. And according to the traffic video features, in-depth study of the integrated vehicle color, shape and other characteristics of the retrieval technique, and then from the perspective of application, given the retrieval system architecture design, Finally, combining of video hierarchical retrieval requirement design a vehicle retrieval system for the transport sector. In this paper, the main research work is as follows:(1) This paper first introduces the content based video retrieval technology background and the research present situation, the basic concept and key technology. Research of the key frame extraction technology, and according to the traffic surveillance video features, adopted a based on the motion area of the key frame extraction method. And combined with the virtual detection line technology on the method to make further improvement.(2) This paper studies the vehicle pretreatment technology and algorithm, and through the experimental method, to find suitable for traffic surveillance vehicle feature preprocessing algorithm, for the follow-up vehicle feature extraction to provide support.(3) Focus on the vehicle feature extraction technology, and according to the traffic surveillance vehicle characteristics, focusing on the vehicle color, shape and geometrical feature extraction method.(4) Study on the retrieval method using multiple features. And the use of hierarchical retrieval methods, the use of geometrical features initial retrieval, integrated color and edge features were the second retrieval methods on retrieval processes are optimized. The two layer retrieves complement each other; the efficiency is improved, but also ensures the retrieval precision.(5) According to the above study method, preliminary design and implementation of a prototype vehicle retrieval system, the prototype system provides several different retrieval methods, convenient user retrieval demand expression.