NORMA eResearch @NCI Library

Real-time Motorcyclists Helmet Detection and Vehicle License Plate Extraction using Deep Learning Techniques

Kanakaraj, Sushaant (2021) Real-time Motorcyclists Helmet Detection and Vehicle License Plate Extraction using Deep Learning Techniques. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (933kB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


On an average 6 two-wheeler riders encounter a fatal accident every hour in India. According to WHO, 42% of those lives could be saved just by ensuring correct usage of Helmets by the rider and the pillion rider. Identifying and penalizing the riders without Helmet could significantly improve this situation. However, it is impractical to employ traffic police personnel on every road to check compliance. This project addresses this issue by identifying non-helmeted motorcyclists in real-time from video footage sourced from a traffic surveillance camera. Licence plate image will be detected and the alphanumeric characters had been extracted which can be later used for identifying and penalizing the rider. This study uses two different state-of-the-art object detection algorithms that have been never tried before in other researches for Helmet detection, YoloV4-Darknet and YoloV5s, which will be compared to find the best model suited for this application. The License plate detection is achieved by the MobileNetV2 FPN lite model and the alphanumeric characters are extracted using EasyOCR. The YoloV4-Darknet model has achieved a mAP of 67.67% and the YoloV5s model has achieved a precision of 51.06%. The MobileNetV2 model achieved a confidence score of 100% for detecting the license plates. It is concluded that the YoloV5s model better suits this application due to faster training times, lightweight architecture, easier prototyping and deployment. The precision scores can be improved by having access to better GPU and more iterations on the training time of the model.

Item Type: Thesis (Masters)
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QA Mathematics > Computer software
T Technology > T Technology (General) > Information Technology > Computer software
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Clara Chan
Date Deposited: 03 Dec 2021 18:09
Last Modified: 03 Dec 2021 18:09

Actions (login required)

View Item View Item