• sensor fusion
• real time
• car to car communication
• camera
• algorithm
• adas
• target tracking
• vehicles

The aim of this thesis is to propose an innovative algorithm to unambiguously assign detected vehicles from Car to Car Communication and an on-board camera. On-board surround sensors, such as cameras and radars, are used in Advanced Driving Assistance Systems applications to improve the driver safety and comfort. In the meantime, Car to Car Communication systems are in the deployment phase and have been tested in huge test fields. In order to compensate the weaknesses and benefit from the strengths of both systems, their information could be fused. In this context, one major challenge is the unambiguous assignment of detected vehicles from different sources, which is still an open research topic and the major target of this thesis. In the first part of the work a comprehensive analysis of the state of the art technologies and applications is performed. Then, the main challenges faced during the design are described in detail and a solution to each problem is proposed. In order to obtain the temporal alignment, the existing hardware setup is modified and specific libraries are written to directly access the data published from the camera on the CAN bus. Concerning the spatial alignment, the data from each source are preprocessed and converted into the Universal Transverse Mercator coordinate system, which is used as the common reference frame. The tracks of the vehicles are initiated and maintained using the Kalman filter, to update the position more frequently, remove the outliers and smooth the trajectory. Finally, the tracks are compared using the Euclidean distance and the vehicle assignment is performed. The algorithm is developed and tested in Matlab using various metrics to evaluate the results. Recorded data from real driving scenarios are used, hence specific tools for data acquisition and storage are deployed. Finally, to demonstrate its capabilities in real time and to offer a convenient environment for further research, the algorithm is implemented inside an in-car system programmed in Java using an OSGi framework. The results obtained are interesting and promising and, from a spatial point of view, they show already a successful matching of vehicles. Compared with the solutions proposed in literature, the demonstrator developed in this thesis is really innovative, and represents the first step towards a real world application running in real time inside cars. Concrete ideas and possible solutions for the future work are given as well. Overall, this thesis is a useful contribution to active safety and autonomous driving applications based on sensor fusion and a good reference for further research on the topic.