Research On License Plate Recognition System

License plate recognition technology is an important part of intelligent transportation.It is an image processing technology that uses license plates to recognize vehicles.With the continuous promotion and deepening of intelligent transportation,intelligent transportation technology covers most fields such as electronic police,highway checkpoints,criminal investigation and so on.In the future,with the continuous improvement of scientific and technological level and the continuous subdivision of intelligent transportation technology application scenarios,such as smart construction site,smart gas station,motorcycle and other small motor vehicle recognition and other application scenarios,license plate recognition technology will usher in broader application space and more application challenges.Firstly,according to the actual situation,this thesis compares the research results of license plate recognition technology at home and abroad in recent years,and gives the specific analysis of license plate recognition technology at home and abroad and the evaluation index of license plate recognition technology.Then,this thesis systematically expounds the license plate recognition algorithms based on segmentation and free segmentation in license plate recognition technology.In the stage of license plate detection,the main features of license plate image are analyzed,and the detection algorithm combining multiple features in license plate detection is deeply analyzed.This thesis expounds the mathematical principle of randon transform,and carries out the tilt correction experiment on the license plate with an inclination of more than 10 degrees by using randon transform.The commonly used character segmentation methods include character segmentation and projection segmentation.In the stage of character recognition,the recognition algorithms based on neural network and pattern recognition algorithms based on the combination of multiple features are compared and analyzed,and the recognition characteristics,advantages and disadvantages of these methods are studied.Finally,aiming at the problem of false recognition of similar characters in license plate image,according to the weak sensitivity of template matching method to noise and the description of character shape by directional gradient histogram feature,this thesis proposes a license plate recognition method based on template matching method and improved local directional gradient histogram feature.The directional gradient histogram feature can maintain good invariance to the geometric and optical deformation of the image.The improved method determines the optimization parameters of the hog feature block of the character image,reduces the dimension of the generated hog feature descriptor,and avoids the problem that the extracted character feature information cannot match the features in the template library.This method not only ensures the recognition speed,but also improves the recognition rate.