Real-time Detection Models for Surrounded Environmental Objects in Highly Automated Vehicles and Inland Vessels

Since the early 1920s, the discourse surrounding highly automated vehicles has evolved significantly, from the initial concept of "phantom autos" to contemporary advancements in driverless technology. This thesis explores the critical role of object detection systems in enhancing the safety, efficiency, and reliability of autonomous vehicles and waterway inland vessels. With the advent of advanced sensors such as lidar, radar, and cameras, effective object detection has become essential for real-time navigation and collision avoidance. The research investigates state-of-the-art object detection techniques tailored to the unique challenges faced by autonomous systems operating under diverse environmental conditions. It evaluates novel algorithms, assesses their performance through extensive simulations and real-world experiments, and identifies limitations that must be addressed for practical implementation. Key objectives include analyzing the variability of objects in different contexts, overcoming sensor limitations, and developing robust sensor fusion strategies to enhance detection accuracy. This work not only contributes to the existing literature by identifying research gaps but also provides actionable insights for the integration of advanced object detection technologies in autonomous transportation. The thesis investigates and develops state-of-the-art object detection techniques tailored to the unique challenges faced by autonomous systems operating under diverse environmental conditions. While these methods may contribute to reducing accidents and improving operational efficiency, this is not explicitly proven in the scope of this work. Instead, the focus is on enhancing environmental perception, which supports the automation of vehicles and inland vessels. This enables safer and more efficient integration into modern transportation networks, though the potential benefits require further investigation beyond the findings presented here.