Automated Helmet
Detection, License Plate
Recognition, and eChallan
Integration for Improved
Road Safety
Authors: Adarsh Kumar, Dr. Deepak Kumar Singh
Institution: Indian Institute of Information
Technology Lucknow
Introduction
• Road safety and law enforcement are
critical in reducing traffic accidents and
ensuring regulatory compliance.
• India faces a significant challenge with
road safety, reporting 461,000 road
accidents and 168,000 deaths in 2022.
• This study aims to develop an automated
system for helmet detection and number
plate recognition to improve road safety.
Objectives
• Detect helmet usage and
Develop an recognize number plates using
Automated YOLOv5 and OCR.
System: • Generate and issue e-challans
automatically.
• Enhance road safety and traffic
Real-Time
law compliance in urban
Enforcement: environments.
Helmet Detection
• Early Methods:
• Manual monitoring or simple image processing techniques, not scalable for
real-time applications.
• Machine Learning Approaches:
Literature • Dahiya et al. (2016):
• Applied a convolutional neural network (CNN) to track motorcyclists with
Review - and without helmets.
• Training the CNN to accurately distinguish between helmets and non-
Helmet
helmets is time-consuming.
• Chiu et al. (2017):
• Developed a system that uses computer-based programs to locate and
Detection separate motorcycles from other vehicles.
• Utilized edge detection for helmet identification.
• Moon et al. (2018):
• Proposed a multi-object tracking algorithm adaptable for bike riding
situations.
• Tao et al. (2019):
• Presented a global motion estimation technique for extracting moving
objects.
• Wen et al. (2013):
• Employed circular arc detection based on the Hough transform to
identify helmets.
• Method limited by reliance on geometric features, leading to false
positives.
Number Plate Recognition:
• Traditional Methods:
• Template matching and edge detection, susceptible to lighting and plate design
Literature
variations.
• Deep Learning Advancements:
• Gnanaprakash et al. (2021):
Review - • Used deep learning for automatic number plate recognition in real-time
applications.
Number Plate • Achieved high accuracy even with moving objects.
• Silva et al. (2014):
Recognition • Proposed a helmet acquisition model using image descriptors and classifiers.
• Method limited by reliance on circular Hough transform, causing incorrect head and
and E-Challan
helmet separation.
E-Challan Generation:
Generation • Early Implementations:
• Image processing and manual data entry.
• Machine Learning Integration:
• Karwal et al. (2015):
• Integrated machine learning for automated detection and fine generation.
• Current Systems:
• YOLOv5 for helmet and number plate detection.
• OCR for text extraction.
• Automated e-challan generation streamlines the process, reducing human
intervention and increasing efficiency.
A. Data Collection
o Data was collected from the college campus and manually annotated.
o Used public datasets and synthetic data augmentation techniques to enhance robustness.
B. Model Training
o YOLOv5 selected for its speed and accuracy.
Methodology
o Hyperparameters optimized, training conducted with separate classes for helmets, motorcycles, and
number plates.
C. Number Plate Recognition with OCR:
o Tesseract OCR engine used for text extraction from number plates.
o Ensured accurate alphanumeric character recognition.
D. System Integration
o Developed a processing pipeline from detection to e-challan generation.
o Utilized SMTP and Twilio API services for notification delivery.
E. Performance Evaluation
o Metrics used: Precision, Recall, F1-score, OCR accuracy, processing speed.
o End-to-end evaluation of system performance in real-world scenarios.
System Architecture
Key Points:
• YOLOv5 architecture: Backbone for
feature extraction, neck for feature
fusion, and head for object
detection.
• Employed Generalized Intersection
over Union (GIoU) Loss Function for
better accuracy. Diagram: System flow from raw traffic footage to
e-challan generation.
• Used Leaky ReLU and sigmoid
activation functions.
Results
High detection accuracy
for helmets, motorcycles,
and number plates
(Precision: 95%, Recall:
92% for helmets).
OCR accuracy: Character
recognition at 89%, plate
recognition at 85%.
The loss and mAP (mean average precision) chart in YOLO
Successful end-to-end e-
challan generation and
delivery.
Output
EasyOCR
Extract Text
from Image
Output
Sample SMS & Mail alert: RazorPay payment gateway
Discussion
• The system effectively detects helmetless riders and
recognizes number plates with high accuracy.
• Challenges include performance in poor lighting and
occlusions.
• Recommendations for future work: Improve
robustness to adverse conditions, develop custom
OCR models, integrate with broader traffic systems,
and address privacy concerns.
Conclusion
• The study presents a
comprehensive system for
automated traffic law
enforcement.
• Demonstrates significant
potential to enhance road
safety and compliance.
• Future improvements can
further increase reliability
and applicability in
diverse environment
Thank You
References
• Dahiya, et al., “Automatic detection of bike riders without helmet using surveillance videos in
real-time,” International Journal of Computer Applications, vol. 150, pp. 15–20, 2016.
• Chiu, et al., “Motorcycle detection and tracking system with occlusion segmentation,” IEEE
Transactions on Intelligent Transportation Systems, vol. 18, pp. 1331–1341, 2017.
• Moon, et al., “A multi-object tracking algorithm for use in sporting events,” Journal of Artificial
Intelligence Research, vol. 62, pp. 399–420, 2018.
• Tao, et al., “Global motion estimation technique for extraction of moving objects,” Pattern
Recognition Letters, vol. 125, pp. 71–79, 2019.
• Wen, et al., “The safety helmet detection for ATM’s surveillance system via the modified
Hough transform,” Pattern Recognition, vol. 46, pp. 156–162, 2013.
• Gnanaprakash, et al., “Automatic number plate recognition using deep learning,” Computer
Vision and Image Understanding, vol. 202, p. 103087, 2021.
• Silva, et al., “Helmet detection on motorcyclists using image descriptors and classifiers,” IEEE
Transactions on Intelligent Transportation Systems, vol. 15, pp. 996–1003, 2014.
• Karwal, et al., “Vehicle number plate detection for Indian vehicles,” International Journal of
Computer Applications, vol. 120, pp. 22–28, 2015.
• Thombre, S. S., Wazalwar, S. S., Tarekar, R., Ramgirwar, S., Zulkantiwar, S., and Dubey, S.,
“Automated traffic rule violation detection with e-challan generation for smart societies,” pp.
65–81, 2022.
• Zhu, L., Geng, X., Li, Z., and Liu, C., “Improving YOLOv5 with attention mechanism for
detecting boulders from planetary images,” Remote Sensing, vol. 13, p. 3776, 09 2021.
• Padalia, D., “Detection and number plate recognition of non-helmeted motorcyclists using
YOLO,” 08 2022.
