Public defence in Mechanical Engineering, M.Sc Nilusha Jayawickrama

Title of the thesis: Machine Vision Analysis of Vehicle Interiors and Surrounding Road Users in Intelligent Transportation

Thesis defender: NIlusha Jayawickrama
Opponent: Prof. Anton Rassolkin, Tallinn University of Technology, Estonia 
Custos: Prof. Kari Tammi,Aalto University School of Engineering

In the near future, our day-to-day travel will be transformed by intelligent transportation systems. Technologies like connected vehicles, self-driving cars, and smart mobility services are developing rapidly. These innovations promise to make travel safer, more efficient, and more accessible. This thesis delves into utilizing machine learning techniques to address two core practical challenges within this domain: how to determine if vehicle seating areas—especially in car-sharing services—are consistently clean, and how to ensure that vehicles with autonomous capabilities can reliably identify the most relevant road users in their surroundings.

The study explored how camera-based computer vision combined with well-matched prediction models can help vehicles increase the awareness of both their interior and exterior environments. Inside the cabin, the methods focused on automatically detecting forgotten personal belongings or various items of trash in a timely manner, aiming to primarily keep shared vehicle fleets pleasant and ready for the next passenger. Outside the vehicle, the research developed models that allow vehicles to recognize road users, such as vehicles and pedestrians, which are most relevant for safe driving decisions—based on a combined detection of how close they are to the vehicle and their manner of motion. These exterior detection capabilities were evaluated across different driving settings, including urban, motorway, and suburban roads.

The methodologies were designed with practical use in mind, with the goal of contributing towards the wider adoption and development of car-sharing services and autonomous vehicles. The results show that the developed perception models performed accurately and reliably even under real-world complexities, reducing the need for manual vehicle interior inspections and improving the ability of autonomous vehicles to understand their surroundings. Limitations of the implemented models were openly analyzed to identify and present clearly key improvements which still remain. The results of the thesis can be applied for enhancing vehicles’ cabin monitoring and exterior perception systems, thereby supporting the broader development of intelligent transportation. Importantly, the research findings and implementations have been published openly, providing a solid foundation for future research.

Keywords: Computer Vision, Intelligent Transportation

Thesis available for public display 7 days prior to the defence at Aaltodoc. 

Contact information: 
Email: nilusha.jayawickrama@aalto.fi 
Linkedin: https://www.linkedin.com/in/nilushacj/ 
Research group: https://www.aalto.fi/en/department-of-energy-and-mechanical-engineering/autonomy-mobility-lab