Study Of Vehicle Type Automatic Recognition System Based On Neural Network

With continuous development of our national economy, the transportation management is getting more and more important in the positions of people's economy and social activity. We also put forward more and more new request on the quality and the Level of transportation management. The current transportation is mainly a manual management in our country, we need to devote a great deal of manpower, material resources and financial powers, and this kind of management can not be adapted to the modern request any more. Therefore, it is important that how to carry out the modernization management to highway, city transportation, and toll stations for each kind of vehicle and the large parking lot waits.For this present condition, the author researches some foundational calculated ways to the vehicle type and identified technique, the main research results are as follows:Firstly, the author discusses an investigative background for identifying car and its important meaning, introducing the domestic and international present research condition in detail and our country's current method of identifying vehicle, pointing out the limitations and shortages of a few existent methods for identifying vehicle. It discusses the academic foundation of artificial nerve and basic structure of BP network, also does the research comparisons to the different kinds of typical nerve network, and introduces BP nerve's structure, characteristics and description of the mathematic way. At the same time, it puts forward the improvement method, making use of the BP nerve network which can enlarge the knowledge, carrying on the independent study, and copying characteristic of person's brain to deal with the comprehensive problems. The method is ready for identifying the follow-up vehicle type.Secondly, this analyzes various shapes of vehicles according to the national standard, summing up the general shape "å·¥" of a vehicle by abstract. Parameters which make from the vehicle crest medium cent of perpendicular line next the bottom grow in front and back comparison etc, is picked up by a great deal of data analysis and handlings. The thesis carries on the improvement to identified technique of the existing vehicle type, deducing the theory for identifying a vehicle type. The theory sets up many layers of reasonable BP nerve network for identifying vehicle, giving the structure of the network in length and breadth, making sample exercises for the network, and adjusting the power value and valves value of the network, which attains superior effect on identification lastly, and applies the BP's nerve network to the vehicle type for identifying the system. The method synthesizes the related calculated ways of the nerve network, identified mode etc, carrying through the general identification to the target outline of vehicle, and comes to a higher identified rate.Finally, the usage of mathematics tool MATLAB 7.0 carries out the calculated way, and discusses imitated result, making use of the Visual C ++ to open up the identified system of vehicle type which bases on the operated system of Windows series at the same time, realizing identified method and the calculated way we mentions above, obtaining the favorable actual result.