License Number Plate Recognition Using Convolution Neural Network

Automation number plate recognition (ANPR) is used to extract the license plate information of vehicle. In many countries it is used for traffic control and toll tax collection without human intervention. There are many ANPR system methods that have been proposed, and still researchers are proposing new methods for the same. License plate recognition (LPR)/ANPR is basically divided into three main steps: license plate detection, character segmentation and recognition of characters. The number of vehicles in India is rising day by day, and due to that, accident cases and traffic rule violation are also increasing. In this paper, two methods are proposed to detect the license plate number. In the first method, segmentation and matching of character with template is used to find the number plate. MATLAB is used for this approach. In second method, convolution neural network (CNN) with ReLU activation function is used to recognize the number plate using Python. In both the approaches, license plate digits and characters are analyzed. We are getting 80% accuracy with first approach and 96% accuracy with second approach. In our dataset totally 1000 license plate images are there.