Real-time vehicle detection and speed estimation system using Raspberry Pi and camera module

Abstract

In the era of intelligent transportation systems, real-time vehicle detection and distance estimation play a crucial role in enhancing road safety and traffic efficiency. This study proposes a low-cost, real-time system that integrates you only look once–version 8 (YOLOv8)-based deep learning for vehicle detection with monocular vision techniques for distance estimation, implemented on a Raspberry Pi embedded platform. The objective is to provide a scalable, affordable solution for traffic monitoring and collision avoidance in resource-constrained environments. The methodology involves using a camera module connected to Raspberry Pi for live video capture, YOLOv8 for object detection, and a calibrated monocular distance estimation algorithm based on bounding box dimensions and known vehicle sizes. Experimental results show that the system achieves over 90% detection accuracy under standard lighting conditions and maintains a distance estimation error below 10% for vehicles within 15 meters. The model processes video frames in real time (~0.17 seconds per frame), proving its effectiveness for embedded deployment. In conclusion, the proposed system offers a robust, power-efficient alternative to high-cost light detection and ranging (LiDAR) or stereo vision systems. Its modular design supports future enhancements such as speed estimation or multi-camera integration, making it highly relevant for smart city applications and low-cost vehicular safety systems.