Overview
A computer-assisted image recognition of a warning system in a vehicle has to react potentially much faster than a human being. Hence, there are clear consequences for the nature of the selection of concepts and procedures deployed for problem resolution.
Goal of the second project phase was to develop a demonstrator on the basis of the results of the first project phase, which can be applied in a vehicle and meets the requirements for real-time and detection reliability. At the same time, such a device should be equipped with a human-machine-interface, which allows for visually informing the driver about the existence of pedestrians without distracting him.
During the development, a time-based sample analysis was forgone but a system with the methodological basis of artificial neural networks in combination with a pre-segmentation via circular hough-transformation was developed, which can detect pedestrians through static images with a high probability quite reliably.
Funding
Results
After the completion of the second project phase, a fully functioning demonstration model is available, which can be applied in a vehicle to inform the driver or increase his attention towards the existence of pedestrians in traffic situations. However, errors can occur. Consequently, the current demonstration status is not practicable to actively intervene in vehicle controls, which though would anyway contradict the definition of objectives of this project.
Findings of the study are published in detail by a final report (German only) which is available online via
https://www.tib.eu/de/suchen/id/TIBKAT%3A351953426/Videobasiertes-Assistenzsystem-zur-Erh%C3%B6hung-der/?tx_tibsearch_search%5Bsearchspace%5D=tn