| With the rapid development of China’s economy,urban motor vehicles have increased rapidly,and various types of public security and criminal cases related to vehicles have also increased year by year.According to the data,more than 65% of criminal cases are related to vehicles.Under this circumstance,new systems such as bayonet and electric police are newly built,and the vehicle can be identified by the license plate information.However,there are problems such as many errors in vehicle identification and less identification information,especially if the license plate as the unique identification code is artificially When occluding,removing(unlicensed cars)and forging(deck cars),it is difficult to effectively and quickly identify the vehicle.This paper intends to develop and construct a road vehicle secondary analysis system that closely matches the actual work of the first-line police and has the functions of vehicle image acquisition,processing,analysis and application.Based on the needs of public security business,this paper studies the architecture based on GPU clusters,and combines the current mainstream deep learning technology and big data technology to improve the accuracy of vehicle identification by using a large number of vehicle models in the public security agency’s motor vehicle database.By subdividing each function,listing the function points,making a flow chart for each function point,and combining the code and timing diagram,the specific implementation process is described.The data a ggregation layer,data processing layer,data application layer and network transmission layer are respectively constructed,and have five functional modules: front-end image acquisition,vehicle big data,multi-engine processing,vehicle image secondary recognition analysis,and business application.Finally,a system architecture that can be deployed on the public security information network and the video private network to connect with multiple systems to realize vehicle data analysis is formed.By testing the important functional modules of the system in the actual scene,the results show that the system can meet the combined retrieval of public security services,and capture various features of the vehicle in the video image(including license plate,model,logo,body color,security).Belts,sun visors,car slanting objects,etc.are analyzed to realize multiple functions such as vehicle secondary identification,system search,big data research and judgment,inspection and control,and remodeling the vehicle trajectory,further enhancing the public security organs to detect various types of vehicles involved.The ability of the case is expected to become a new professional technology for public security organs to investigate and solve crimes. |
