International Journal for Research in Applied Science and Engineering Technology (IJRASET) is an international peer-reviewed, open-access & multidisciplinary online journal published for the enhancement of research in various disciplines of Applied Science & Engineering Technologies.

Detection of Motor Bicyclist Violating Traffic Rules using Computational Neural Networks

Abstract: This project is based on automatic detection of motorcyclist who are violating traffic rules. Detection of motorcyclist
who are not following traffic rules is quite complicated and laborious. If there are more than one person who are travelling on
the same bike considered as violating traffic rules. Once the traffic violator identified, the number plate of such motorcycle is
extracted. In this paper we discussed about drawbacks of traditional classifier and their drawbacks and to overcome the
drawbacks how our proposed neural network helps. Neural network take decision intelligently as it is a deep learning classifier.
In existing there is use of HOG (Histogram Oriented Gradient) features with SVM (Support Vector Machine) classifier while in
proposed work we used yolo v3 for detection of traffic violators.
Keyword: Traffic violation rules, neural network classifier, Yolo V3 algorithm, Deep Learning
I. INTRODUCTION
The aim of this project is to detect motorcyclists who are violating traffic rules. A person riding a motorcycle without helmet
considered as traffic violator. More than one person riding on the same bike are also considered as traffic violators. After finding the
traffic violator the number plate of his motorbike is extracted. In this project we are folloeing two approaches, traditional classifier
based approach and the second is neural network approach.
1) Neural Network Approach: (yolo) we have used yolov3 to detect person, motorbike, and helmet. If a person is found riding a
motorbike without helmet, then the number plate of the motorbike is extracted using openalpr.
2) Traditional Classifier Approach: ( HOG+SVM) we have used SVM classifier to detect helmet in image. The HOG features of
helmet dataset and non-helmet dataset are extracted and the features are used to train SVM classifier. Then the classifier is
tested to detect helmet in the image. Finally, the number plate of the traffic violator is extracted using openalpr.