In an increasing number of cars, the driver is supported by Advanced Driver Assistance Systems (ADAS). In particular camera based ADAS are a key component for further improvements in safety and driving comfort. While imaging sensors are performing well under good weather conditions, their efficiency suffers under adverse environmental influences such as heavy rain, fog or snow. To handle such optical threats and to estimate information quality of cameras in order to warn the assistance system of possible critical working conditions, a self-diagnosis mechanism is of great importance for reliable operation of an optical ADAS.In this paper an approach of camera based fog detection as part of a self-diagnosis mechanism for ADAS based on the blurring effect of fog is presented.The encouraging results of our experiments have shown that the presented approach of analysing the power spectrum slope (PSS) of a small image block in close proximity to the vanishing point enables a fast discrimination of street scenes with and without fog.
This paper shows an initial approach of a system architecture for a wireless sensor network (WSN), addressing the maritime domain. Novel architectures, technologies and prototypes for WSNs have been in focus of researchers and economists for several years. The proposed system architecture is focused on maritime automation -the construction and maintenance of offshore wind farms with their specific requirements. Harsh environmental conditions impede the performance of wireless technologies in onshore and offshore applications, caused for instance by many metal components, wayless terrain, often nonline-of sight (NLOS) connections between mobile motes, and dynamic ground motion for example onto a jack up ship for offshore construction work.Furthermore, many approaches for WSN are addressing either communication or localization networks. The proposed system architecture covers both at the same time, based on Ultra Wide Band radio technology (UWB). UWB allows robust distance measurements and communication, in particular in harsh environments.
For the introduction of new automated driving functions, the systems need to be verified extensively. A scenario-driven approach has become an accepted method for this task. But to verify the functionality of an automated vehicle in the simulation in a certain scenario such as a lane change, characteristics of scenarios need to be identified. This, however, requires to extract lane-change from real-world drivings accurately. For that purpose, this work proposes a novel framework for lane-change identification by combining multiple unsupervised learning methods. To represent various types of lane changes, the maneuver is split up into primitive driving actions with an Hidden Markov Model (HMM) and Divisive Hierarchical Clustering (DHC). Based on them, lane change maneuvers are identifier using Dynamic Time Warping (DTW). The presented framework is evaluated with a realworld test drive and compared to other baseline methods. With a F1 score of 98.01% in lane-change identification, the presented approach outperforms the other approaches.
In this paper, we introduce a novel shadow removal technique that produces a shadow-free scene. There have been few studies concerning shadow removal, and the existing approaches cannot perfectly restore the original background patterns after removing shadows. With an acceptable number of differently illuminated images, the proposed algorithm simulates an artificial infinite illuminant plane over the field of view. By employing the offset reduction technique, the constancy of the brightness is also reliably guaranteed. Finally, a shadowless image without loss of textural details is obtained without any region extraction phase. Experimental results show that the method could successfully remove all of the visible shadows. The benefits of the proposed algorithm compared to the conventional shadow detection algorithms are the lower computational costs and the improved reliability.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.