“AI Based RTO Safety Compliance Tracking System for Agricultural Vehicles”
Shital Gaikwad
Asst. Prof., DKTE Society's Textile & Engineering Institute (An Empowered Autonomous Institute), Ichalkaranji, Maharashtra, India.
Email: ssgaikwad2712@gmail.com
Parth Naykude, Rishabh Mehta, Pruthviraj Mudekar, Avishkar Nalawade, Nikita Patil
Department of Computer Science and Engineering (AI)
DKTE Society’s Textile & Engineering Institute (An Empowered Autonomous Institute), India
Email: {parthnaykude111222,rishabhmehta790,pruthvirajmudekar23,avishkarnalawade950,nikitapatil172004}@gmail.com
Abstract—Agricultural vehicles such as tractors and trolleys play a vital role in rural transportation and farming operations. However, many of these vehicles frequently violate safety regulations mandated by the Regional Transport Office (RTO), including the absence of rear reflectors and mandatory red warning cloth indicators. Such violations significantly increase the risk of road accidents, particularly during nighttime and low-visibility conditions. Manual monitoring in rural and semi-urban areas is challenging due to limited manpower and vast geographical coverage, creating the need for an automated compliance monitoring system.
This paper proposes an AI-based RTO Safety Compliance Tracking System that leverages computer vision and deep learning techniques to automate detection, verification, and reporting of safety violations. The framework integrates YOLOv8 for real-time tractor detection, a MobileNetV2-based convolutional neural network for safety compliance verification, and Tesseract Optical Character Recognition (OCR) for automatic license plate extraction. A duplicate detection mechanism using Structural Similarity Index (SSIM) and ORB feature matching prevents redundant reporting of the same vehicle.
The system processes live CCTV feeds or recorded videos, generates annotated detection outputs, stores compliance data in a structured database, and provides real-time visualization through a Streamlit-based dashboard. Automated violation reports containing vehicle image, license number, timestamp, and compliance status are generated for RTO authorities. Experimental evaluation indicates promising detection accuracy and near real-time performance, making the system suitable for practical deployment in rural safety monitoring environments.
