Sonstige
Filtern
Dokumenttyp
- Konferenzveröffentlichung (3) (entfernen)
Sprache
- Englisch (3) (entfernen)
Schlagworte
- Camera (3) (entfernen)
Institut
- Sonstige (3)
- Abteilung Straßenbautechnik (1)
The changed focus in vehicle safety technology from secondary to primary safety systems need to evolve new methods to investigate accidents, high critical, critical and normal driving situations. Current Naturalistic Driving Studies mostly use vehicles that are highly equipped with additional measuring devices, video cameras, recording technology, and sensors. These equipped fleets are very expensive regarding the setup and administration of the study. Due to the great rarity of crashes it is additionally necessary to have a high distribution and a homogeneous distribution of subject groups. At the end all these facts are leading to a very expensive study with a manageable number of data. Smartphones are becoming more and more popular not only for younger people. Contrary to traditional mobile phones they are mostly equipped with sensors for acceleration and yaw rates, GPS modules as well as cameras in high definition resolution. Additionally they have high-performance processors that enable the execution of CPU-intensive tools directly on the phone. The wide distribution of these smartphones enables researchers to get high numbers of users for such studies. The paper shows and demonstrates a software app for smartphones that is able to record different driving situations up to crashes. Therefore all relevant parameter from the sensors, camera and GPS device are saved for a given duration if the event was triggered. The complete configuration is independently adjustable to the relevant driver and all events were sent automatically to the research institute for a further process. Direct after the event, interviews with the driver can be done and important data regarding the event itself are documented. The presentation shows the methodology and gives a demonstration of the working progress as well as first results and examples of the current study. In the discussion the advantages of this method will be discussed and compared with the disadvantages. The paper shows an alternative method to investigate real accident and incident data. This method is thereby highly cost efficient and comparable with existing methods for benefit estimation.
The Traffic Accident Research Institute at University of Technology Dresden investigates about 1,000 accidents annually in the area around and in Dresden. These datasets have been summarized and evaluated in the GIDAS (German Accident In-Depth Study) project for 13 years. During the project it became apparent that the specific traffic situation of a covert exit of a passenger car and an intersecting two-wheeler involves a high risk potential. This critical situation develops in a large part due to the lack of visibility between the driver and the intersecting bike. In this paper the accident avoidance potential of front camera systems with lateral field of view, which allows the driver to have an indirect sight into the crossing street area will be presented.
Road condition acquisition and assessment are the key to guarantee their permanent availability. In order to maintain a country's whole road network, millions of high-resolution images have to be analyzed annually. Currently, this requires cost and time excessive manual labor. We aim to automate this process to a high degree by applying deep neural networks. Such networks need a lot of data to be trained successfully, which are not publicly available at the moment. In this paper, we present the GAPs dataset, which is the first freely available pavement distress dataset of a size, large enough to train high-performing deep neural networks. It provides high quality images, recorded by a standardized process fulfilling German federal regulations, and detailed distress annotations. For the first time, this enables a fair comparison of research in this field. Furthermore, we present a first evaluation of the state of the art in pavement distress detection and an analysis of the effectiveness of state of the art regularization techniques on this dataset.