Vehicle-Logo Recognition Method Based On ADABOOST And BP

With the continuous increase of China’s automobile, regardless of the number of motor vehicles or small cars are in rapid growth in a year, and Intelligent management system (ITS) becomes very important. And vehicle identification technology as an important subject, the main research significance is accuracy of identification, and it has good robustness in a variety of environments of the design algorithm. The subject of this paper is Vehicle-Logo Recognition Method Based On ADABOOST and BP.Car logo recognition technology is an important issue based on digital image processing, pattern recognition and computer vision. Car logo recognition system includes image acquisition, pre-processing, license plate location, logo positioning and vehicle logo recognition.The main contents are as follows:Firstly, it discusses the advantages and disadvantages of vehicle identification technology. And then it analyzes the significance, the research status and development trends of computer vision techniques.Secondly, it designs the pre-processing program of the standard car license plate image (image transformation, denoising filtering, morphological operations, etc.), enhancing license plate logo image, highlighting their details, enhancing edges, and then locates the license plate’s location by calculating horizontal and vertical direction of projection, positioning the target position of the bus. And then it uses a different recognition algorithm classifies the car standard,detected and identified by extracting the subject of car features, and it enhances the degree of identification other car automation.Thirdly, it uses the improved ADABOOST algorithm based on vehicle logo recognition technology. It trains weak classifier to strong classifier, and it finally recognizes the non-standard car and standard car. It classifies different kinds of car logo by BP algorithm, the algorithm will make a contrast between the desired output and the actual output, iterate constantly updated weights, and ultimately get thecorrect classification.Finally, it develops the human-computer interaction software by using Visual Studio 2012 integrated development environment and the open source library developed opencv2.4.4, which is used in image acquisition platform-independent environment, pre-processing image automatically, plating license, and logo recognition. The experimental results show that the system has good detection accuracy.