3D Space Object and Road Detection for Autonomous Vehicles Using Monocular Camera Images and Deep Learning Algorithms

This study develops a road-detection and obstacle-avoidance system for autonomous vehicles using a monocular camera and deep learning technology. The system is specifically designed to detect roads and obstacles in real time, thereby facilitating the identification of safe and efficient driving paths. This system captures real-time images with a monocular camera, and deep learning models for object detection and semantic segmentation are developed and implemented. Subsequently, a 3D map is generated based on the detected information using a pinhole camera model. The objective of this study is to develop a cost-effective and streamlined monocular camera-based system applicable to autonomous vehicles, enhancing their safety, reliability, and emergency response capabilities. This system aims to replace or supplement binocular camera systems, thereby reinforcing vehicle safety and providing a technological foundation for using monocular systems as viable alternatives or complements. © ICROS 2024.