Research On Key Technology Of Auto License Plate Recognition

License Plate Recognition(LPR)System is an important part of Intelligent Transportation System(ITS),and has great effect on ITS.LPR main include four parts: license plate location,license plate tilt correction,character segment and character recognition.The first three parts have been studied in the paper.The following are the specific work.Aiming to cope with the difficult and long time of license plate location,a license plate location algorithm based on plate background and character's color feature,and a license plate location based on edge detection and improved Harris corner detection are proposed.License plate location is finished by using new color model and improved Canny edge detection in the first method.In the second method,the location is completed by utilizing raised horizontal and certical edge detection respectively and improved Harris corner detection.The algorithm is effective and can be widely used.A car sticker algorithm based on maximum stable extreme region(MSER)is proposed for solving the effect of sticker.The sticker is detected by using improved MSER detection method.The algorithm is simple and has higher accuracy.Aiming to poor robustness and light sensitivity,a tilt correction algorithm based on the character median line,and a tilt correction algorithm based on color model and corner detection are proposed.In the first method,tilt correction is finished by using projection method,character median line and the least-squares method.Experiments show that the proposed algorithm is simple,has low error ratio,good robustness against noise and deformation.In the second method,tilt correction is completed by using HSI color model and improved Harris corner detection.Experiments show that the algorithm is effective to different light,but it's sensitive to noise.Aiming to low spilt rate,an image segment algorithm based on exponential kernel function is proposed.Firstly,according to the fast convergence character of exponential function,the efficiency is improved.Secondly,the accuracy is more precise by modifying the energy function of the CV model.Finally,introducing level set function avoids the re-initialization.Compared with the CV model,the CV model based on exponential kernel function has higher segmentation accuracy and stronger anti-noise ability,and needs less iteration times and running time.