Research On Civil License Plate Recognition Under Complex Background

The license plate recognition system is the core driving part of the construction of modern urban intelligent transportation.The system includes color space,image grayscale analysis,image denoising,perspective transformation and affine transformation,eolor and character segmentation,image binary processing,mathematical morphology processing and other operation methods.The traditional method of license plate location is disturbed by dim,strong noise,backlight,overexposure,and complicated background.When the positioning based on the complete license plate is used,the positioning result may increase or decrease some edge information,which results in incomplete or excessive location results.After reviewing a large number of experiments of the same type,this thesis summarizes a recognition method that first roughly locates the license plate and then accurately locates the license plate.The effect of shadow on the color of the license plate can be solved to some extent.lt is also possible to avoid the situation where the blue car's body area is inseparable from the license plate.Moreover,the algorithm can locate the license plate area under a complex background.The main work of this thesis is as follows:(1)Introduction.Solve the practical problems of this topie by reading a large amount of relevant literature.Taking the implementation of its own application as the starting point,integrating relevant theories,analyzing the problem characteristics and difficulties of the subject.Thereby determine the objectives stated in the project and the impact on future practical applications.(2)The image is initially processed to ensure that further positioning and identification are completed.Therefore,it is often necessary to perform related image processing on the collected license plate.Pre-processing of the image requires completion of the mutual transformation of the color space,and the color image is subjected to various operations such as decoloring processing,image smoothing processing,and picture noise reduction.(3)The rough positioning and precise positioning are essential parts of the license plate word/number recognition system.(4)After the license plate is positioned,the next operation is correction work.The non?horizontal license plate is level-processed to meet the needs of word/number recognition,achieving perfect segmentation of license plate characters and providing material for recognition and training.(5)Character recognition based on SVM+HOG theory is the last step to identify each character after completing the above work.(6)Organize and summarize the parts of the research that are prone to problems and mistakes,and prepare for the next deeper research work.