Abstract: Traffic sign detection plays a pivotal role in enhancing road safety and enabling autonomous vehicles. Indian roads, with their unique multilingual signs and dynamic environments, present significant challenges due to diverse regional languages, fonts, and conditions. Addressing these complexities is essential for developing reliable navigation systems for autonomous bots and vehicles. This objective aims to design and implement a real-time traffic sign detection system for Indian roads using the YOLOv5 model deployed on a Raspberry Pi. The system integrates multilingual detection capabilities and dynamically displays decisions in a terminal interface, enabling autonomous bots to operate efficiently in unpredictable environments. The methods in the dataset comprising multilingual Indian traffic signs were collected and annotated. The YOLOv5 model was trained with augmented data to enhance detection accuracy. The trained model was optimized for edge devices using TensorRT and Pytorch. A Raspberry Pi, integrated with a depth camera, processed real-time video streams for detection. Detected signs were mapped to pre-programmed actions, which were displayed in the terminal and executed via bot navigation. The Results of the system achieved high detection accuracy and low latency, even under challenging lighting and weather conditions. Multilingual OCR integration ensured robust detection of diverse traffic signs. Real-world tests demonstrated reliable navigation and responsive action execution. The project's significant contribution to traffic management, road safety, and autonomous vehicle technologies, addressing the unique challenges of multilingual environments. The scalable solution has implications for smart city initiatives and real-time navigation systems on Indian roads.
Keywords: YOLOv5, Autonomous Vehicles, Traffic Sign Detection, Multilingual OCR.
|
DOI:
10.17148/IJARCCE.2024.131202
