Vehicle Number Plate Detection and Recognition Using Deep Neural Networks

Main Article Content

Dr. T. K. S. Rathish Babu, S. Shivani, Sangam Vedha Shree, Banglinti Nagalaxmi


Background:  In recent times, the use of car is growing every day. Hence, following the traffic guidelines are taken into consideration to be the predominant component. If the individual does not obey any rules they need to be punished. The prototype of this paper can be like the police can view and take a look at the details consisting of license, coverage that are stored and maintained with the aid of RTO admin. Police can simply view the details of the given variety plate. The quantity plate of the automobile will be in rectangular shape, the best strategies are used which include Non-Max Suppression for recognising the nice bounding field i.e., detecting the numberplate of the car. Subsequently, on this paper we tackle the trouble of wide variety plate detection in natural scene photos. We propose a unified deep neural network and CNN, which localize the pictures of range plate. In contrast to current device which take number plate detection and popularity as two separate responsibilities and works step by step one after the opposite, our method solves these two tasks together in an unmarried network approach.

Objectives: The main Objective is to detect the number plate and recognise the labels in the number plate using deep neural network approach

Methods: The methods involved are:

1.Feature Extraction: In this method to remove the unwanted data the frame should undergo pre-processing.

2.Model Training: To make the model learn, it needs to be trained on the dataset.

3.Number Plate Detection: Here we detect the number plate of the vehicle from real scene images.

4.Character Recognition: In this method we recognise the characters in the detected number plate of the vehicle.

Conclusions: The jointly trained network for Vehicle Number plate detection and the character recognition of the labels in the number plate is presented. With this community, the automobile range plates can be detected and identified all of sudden in a single forward bypass, with each high accuracy and performance. By using sharing convolutional functions with both detection and reputation network, the version size decreases in large part. The whole community may be about educated cease-to-give up, without intermediate steps like photograph cropping or person separation and segmentation. Within the future, we will enlarge our network to multi-oriented vehicle license plates. Further, with the time evaluation, it is located that take approximately 1/2 of the complete processing time. Consequently, we are able to optimize to accelerate the processing speed.

Article Details