Anti-Noise Detection And Recognition Algorithm For Vehicle License Plate Under Complicated Background

License Plate Recognition(LPR)is one of the vital technology in field of the current intelligent transportation,which is widely applied to vehicle supervising and electronic toll collection in parking lot.this technology relates to these fields including computer vision,digital image processing and pattern recognitions,which is also one of the research subjects applied to promote the level of intelligent transportation system.The thesis is the research of LPR algorithm in complex environment.First of all,we briefly introduce the application and future of LPR in China.In addition,we make an introduction to the basic principle of LPR and recognition in complex environment.Furtherly,we mainly focus on the research of methods about car's license location,character's segmentation and character's recognition.These methods are made up of the algorithm framework of the whole system.On the method of license location,we put forward an algorithm based on the fractal dimension and mathematical morphology.We use fractal self-similar properties of the fractal dimension to refactor the model of the gray image.The multidimensional detector with mathematical morphology is used to calibrate the edge of the license plate image and then we can improve the accuracy of license plate positioning.On the method of character's segmentation,we adopt an algorithm that combines projection analysis and recursive fly back.The vertical projection of the characters used to determine the coordinates of the location of each character,and then we can improve the speed and accuracy of character segmentation.On the method of character's recognition,we use character's projection sequence and margin as structural feature so that make a coarse classification.Then,we make a match with character's formwork to recognition.Firstly,according to the structural features of the character itself and the projection sequence of the character image,the coarse sorting of the classifier is used to filter out the possible sequence of characters,and then the corresponding template is identified by matching with the standard template library.