Research On License Plate Recognition Algorithm Under Complex Conditions

At present,the number of motor vehicles in China is continuously rising,and the scene of collecting license plates is becoming more and more complicated.Therefore,it is very important to improve the universality of license plate recognition technology.Research on license plate recognition technology under complex conditions still has great market value.The specific work is as follows:In terms of license plate positioning,the traditional license plate positioning algorithm is not ideal for the license plate positioning under various complex conditions.The area feature of the license plate locates the license plate area.It can be seen from the test results that this algorithm can successfully locate three different types of license plates under various complex conditions,and the positioning success rate is high.In terms of character segmentation,for the problem that the traditional license plate tilt correction algorithm takes a long time,this paper uses the method based on the orientation operator to measure the offset angle of the license plate,and then uses the rotation method to correct the inclined license plate.Experiments show that the algorithm can improve the efficiency of correction without affecting the correction success rate.In character segmentation,a character segmentation method based on inverse color images is studied.By performing inverse color processing on the image after gradation processing,the part of the characters can be highlighted while irrelevant regions can be removed.This method has a small amount of calculation and high segmentation efficiency.In terms of license plate character recognition,the Le Net-5 network prototype cannot directly solve the problem of mixed recognition.This paper improves the traditional Le Net-5network.Enter the image into two different convolutional networks respectively.This method will prevent overfitting and ensure that the image feature information is not lost during the transfer process.This paper applies it to character recognition algorithms under complex conditions,and the final recognition success rate reaches 91.64%.The results show that the license plate recognition algorithm in this paper can achieve better recognition accuracy under complex conditions and has good universality.