A Deep Learning Based Approach for Automatic Detection of Bike Riders without Helmet

Abstract

The most common type of transportation in India is two-wheeler, where the number of accidents are increasing day by-day. In general, these accidents have occurred due to riding a motorcycle without a helmet. It is very difficult to monitor each and every rider whether they are wearing a helmet or not by a human labour , where as an electronic detection system can do the same kind of work without any human effort. Image processing is a solution for this kind of problem where there are many advancements in recent times. This works with extracting the features and identifying the objects which resides in the images which are taken out from the video surveillance as multiple frames. Convolutional Neural Networks (CNN) or Deep learning techniques are used for image or pattern identification along with Visual Geometry Group (VGG) which is mainly used for object detection. Multiple models of CNN are used to train the images to classify different types of motorcycles and head positions of different riders. If a bike rider is found travelling without a helmet, the image of the number plate of the bike is captured. The number plate is checked with the databases and penalty will be issued. The system uses pure machine learning algorithm for image processing. Identification of the motorcycle can be done in five steps: image capturing, pre-processing of image, finding the errors, image recognition, feature extraction. The usage of machine learning algorithms and object detection techniques will improve the robustness and effectiveness for the detection of riders without helmet.