Researchers at the National Institute of Technology Rourkela (NIT-R) have developed an AI-based Multi-Class Vehicle Detection (MCVD) model and a Light Fusion Bi-Directional Feature Pyramid Network (LFBFPN) to enhance AI traffic management. Led by Dr. Santos Kumar Das from the Electronics & Communication Engineering department, the team designed an intelligent vehicle detection (IVD) system that accurately identifies vehicles in real-time, even under challenging weather and traffic conditions. Their study, published in IEEE Transactions on Intelligent Transportation Systems, highlights how this innovation can support intelligent transportation systems in mixed-traffic environments.
The MCVD model leverages a video de-interlacing network (VDnet) to extract key traffic features and uses LFBFPN for further refinement. This approach improves accuracy while reducing computational complexity, making it more efficient than traditional models. The system was tested using the Heterogeneous Traffic Labelled Dataset (HTLD) and on an Nvidia Jetson TX2 edge computing device, demonstrating high-speed and precise detection even with low-resolution images. Unlike expensive sensor-based adaptive traffic control systems, this AI-driven model is more suitable for developing countries where mixed traffic includes bicycles, rickshaws, and pedestrians.
Looking ahead, the NIT-R team aims to integrate this technology into a traffic control system to optimize road safety and congestion management. They are also exploring its commercialization through a start-up, redefining intelligent transportation systems in India’s dynamic road conditions.