In today's world, withincreasing population number of vehiclesare also increased. Monitoring each andevery vehicle in roadside by manualforce becomes a tedious job. For effective traffic management some solution is required. The traditional system of licenseplate recognition basically relies on themorphological processing of images. The accuracy of recognition is also lower. Inorder to provide an efficient solution forlicense plate position, we propose a new method of license plate recognition. In the license plate position, we use traditional positioning method and Haar Cascade algorithm to detect the licenseplate. In addition with the traditionalidentification system cutting and matching, we use the ability feature extraction of Convolution Neural Network (CNN) to detect the license platedirectly, which avoids the recognitionerror which is caused by the segmentation in the license platerecognition. In this project, we propose a new license plate recognition system. It is divided into two parts, they are licenseplate positioning and characterrecognition.