Role of sensory and cognitive conspicuity in the prevention of collisions between motorcycles and trucks at T-intersections.

Motorcyclists are particularly vulnerable to injury in crashes with heavy vehicles due to substantial differences in vehicle mass, the degree of protection and speed. There is a considerable difference in height between motorcycles and trucks; motorcycles are viewed by truck drivers from downward angles, and shorter distances between them mean steeper downward angles. Hence, we anticipated that the effects of motorcycle conspicuity treatments would be different for truck drivers. Therefore, this study aims to evaluate the effects of motorcycle conspicuity treatments on the identification and detection of motorcycles by truck drivers. Two complementary experiments were performed; the first experiment assessed the impact of motorcycle sensory conspicuity on the ability of un-alerted truck drivers to detect motorcycles, and the second experiment assessed the motorcycle cognitive conspicuity to alerted truck drivers. The sensory conspicuity was measured in terms of motorcycle detection rates by un-alerted truck drivers when they were not anticipating a motorcycle within a realistic driving scene, while the cognitive conspicuity was determined by the time taken by alerted truck drivers to actively search for a motorcycle. In the first experiment, the participants were presented with 10 pictures and were instructed to report the kinds of vehicles that were presented in the pictures. Each picture was shown to the participants for 600ms. In the second experiment, the participants were presented with the same set of pictures and were instructed to respond by clicking the right button on a mouse as soon as they detected a motorcycle in the picture. The results indicate that the motorcycle detection rate increases, and the response time to search for a motorcycle decreases, as the distance between the targeted motorcycle and the viewer decreases. This is true regardless of the type of conspicuity treatment used. The use of daytime running headlights (DRH) was found to increase the detection rate and the identification of a motorcycle by a truck driver at a farther distance, but effect deteriorates as the distance decreases. The results show that the detection rate and the identification of a motorcyclist wearing a black helmet with a reflective sticker increases as the distance between the motorcycle and the truck decreases. We also found that a motorcyclist wearing a white helmet and a white outfit is more identifiable and detectable at both shorter and longer distances. In conclusion, although this study provides evidence that the use of appropriate conspicuity treatments enhances motorcycle conspicuity to truck drivers, we suggest that more attention should be paid to the effect of background environment on motorcycle conspicuity.

Ontology-Based Semantic Image Segmentation Using Mixture Models and Multiple CRFs.

Semantic image segmentation is a fundamental yet challenging problem, which can be viewed as an extension of the conventional object detection with close relation to image segmentation and classification. It aims to partition images into non-overlapping regions that are assigned predefined semantic labels. Most of the existing approaches utilize and integrate low-level local features and high-level contextual cues, which are fed into an inference framework such as, the conditional random field (CRF). However, the lack of meaning in the primitives (i.e., pixels or superpixels) and the cues provides low discriminatory capabilities, since they are rarely object-consistent. Moreover, blind combinations of heterogeneous features and contextual cues exploitation through limited neighborhood relations in the CRFs tend to degrade the labeling performance. This paper proposes an ontology-based semantic image segmentation (OBSIS) approach that jointly models image segmentation and object detection. In particular, a Dirichlet process mixture model transforms the low-level visual space into an intermediate semantic space, which drastically reduces the feature dimensionality. These features are then individually weighed and independently learned within the context, using multiple CRFs. The segmentation of images into object parts is hence reduced to a classification task, where object inference is passed to an ontology model. This model resembles the way by which humans understand the images through the combination of different cues, context models, and rule-based learning of the ontologies. Experimental evaluations using the MSRC-21 and PASCAL VOC'2010 data sets show promising results.