Real-time Vehicle License Plate Detection by Using Convolutional Neural Network Algorithm with Tensorflow

Abstract

Detection of vehicle license plates (VLP) become a challenging issue because of the variability in conditions and the types of the license plate. There are several solutions use stationary cameras with a limited angle and specific resolution and also for a specific license plate type. Unfortunately, license plate detection is a challenging issue when vehicles in the open environments and images are taken from a particular range by low cost cameras. Vehicle license plates can be detected with computer vision technology in real-time video conditions. In this paper, we propose vehicle license plates detection system by using convolutional neural network (CNN) with tensorflow. The research was conduct in three step process such as data pre-processing, training/testing process, and interpretation result. Using CNNs algorithm in tensorflow with 25,000 steps and 8 batches on the training process can produce a training model of vehicle license plates detection with high accuracy around 70-99%.