Helmet detection, gaze tracking among VNIT’s 4 solutions for safe, smart roads

Nagpur: From detecting riders without helmets to tracking where drivers focused their eyes, researchers at Visvesvaraya National Institute of Technology (VNIT) developed four innovative traffic solutions aimed at making Indian roads safer and smarter.VNIT experts shared details of these studies with TOI ahead of the National Science Day on February 28, highlighting how artificial intelligence (AI), machine learning (ML) and low-cost Internet of Things (IoT) devices can transform urban traffic management.
One of the solutions is an automated system that uses CCTV feeds and deep learning models to identify whether a two-wheeler rider is wearing a helmet. Pre-trained neural networks such as MobileNet-SSD first detect motorcycles in traffic footage, after which a region-of-interest method isolates the rider's head for helmet verification. Trained on more than 3,400 images, the system achieved around 79% testing accuracy.Researchers say the technology can enable large-scale, real-time monitoring and support traffic police in improving enforcement consistency while reducing manpower requirements.In another patented development, the research team created an IoT-based Autonomous Traffic Monitoring and Analysis System (ATMAS). Built on an embedded computing platform, the system automatically detects and classifies vehicles, converts image coordinates into real-world coordinates, and estimates speed and acceleration profiles.The system also extracts trajectory patterns and lane behaviour data. Unlike traditional manual traffic surveys, ATMAS continuously generates traffic volume, vehicle mix and speed data that can be stored digitally and uploaded to cloud servers. Such reliable datasets can assist authorities in signal timing optimisation, congestion management, accident analysis and scientific road design.These studies were pioneered by Udit Jain from the Department of Civil Engineering.Addressing the need for affordable congestion monitoring, the third study introduced a low-cost Travel Time and Stream Speed Analyser using ESP32 microcontrollers with Bluetooth capability. Two compact devices installed at entry and exit points of a corridor capture anonymous Bluetooth MAC IDs within a limited range. By matching IDs recorded at both points, the system estimates travel time and stream speed.Field trials showed a penetration rate of 12–13% and a match rate of 65%, with statistical validation confirming consistency with video-based measurements. The portable and energy-efficient device offers an economical way for cities to monitor congestion in real time.The fourth study examines driver behaviour through a wearable IoT-based gaze detection system. The device tracks pupil movement, gaze projection and head pose while integrating deep learning-based lane and road sign detection. Heat maps generated from the data reveal attention patterns and identify risky behaviour linked to distraction or fatigue.Tested on both international datasets and Indian road conditions, including in Nagpur, the system has potential applications in driver monitoring, fatigue alerts and advanced driver assistance systems.Collectively, the four studies demonstrate how indigenous, low-cost technologies can support smarter enforcement, evidence-based planning and improved road safety, offering Indian cities a scalable path towards intelligent transportation systems.