Deep Learning Algorithm based License Plate Detection for Traffic Control

In today's world, we come across various incidents of traffic violations which can be solved with a number of approaches. Riding motorcycles/bikes without a helmet is violating the traffic rules which has led to a drastic increase in the number of road accidents and deaths. The already existing methods requires a lot of time and manpower since the number of violators are large in terms of frequency due to increase in the number of daily bike riders. Hence, a system which would automatically look for non-helmet riders and extract their number on the license plate is important. This paper explains the procedure to read the license plate of the riders who do not wear helmets. In this paper, object detection using neural networks and deep learning in multiple stages is proposed. The objects detected are humans, bikes (two-wheelers) in the first stage, helmet detection in the second and license plate number extraction in the last stage using deep learning algorithms. Results are shown to validate the performance of the proposed method.