No abstract

No abstract

Typical road users appear not to understand retroreflectivity despite nightly exposure to retroreflective materials like road signs. A critical benefit of retroreflective materials is a robustness to changes in entrance angle, the angle at which light strikes the material. This study aims to measure observers’ perceived brightness judgments of surfaces representing three types of reflection (diffuse, specular, and retroreflective) when viewed under manipulations of entrance angle. Perceived brightness will be assessed before and during a demonstration including illumination from a source positioned near the observer’s eyes. Prior to the demonstration, observers are hypothesized to predict specular and retroreflective surfaces will have a consistent brightness despite changes in entrance angle. Seeing the retroreflectivity demo is expected to result in increased perceived brightness of only the retroreflective surfaces in the more extreme entrance angle conditions. Watching visual demonstrations of reflection, however, is expected to produce an enhanced appreciation that retroreflective (but not specular or diffuse) surfaces remain bright despite large changes in entrance angle. This evidence may eventually increase demand for retroreflective markings by vulnerable road users.

Bicyclists risk being involved in collisions with motor vehicles, even during daytime. Thus, bicyclists who ride in daylight must enhance their conspicuity. This study assessed the daytime conspicuity benefits of bicycle taillights using eye tracking technology. Participants were driven along an open-road route while wearing an eye tracker and pressed buttons when they detected and recognized a test bicyclist. Participants encountered the bicyclist displaying one of four taillight configurations, and the distances from which they responded to the test bicyclist were recorded. The results revealed that, after participants first glanced at the bicyclist, a significant amount of time was needed to detect and recognize the bicyclist. Further, seat post-mounted lights displayed with or without lights mounted to the heels of the rider’s shoes provided the greatest conspicuity advantage in terms of recognition. This experiment offers useful insights into the optimal light placement options for bicyclists to enhance their daytime conspicuity.

This paper presents a hierarchical connectivity scheme to implement a totally connected Artificial Neural Network. An ANN chip unit ( ACU) is the basic building block that implements a l&input/output net, and the system is easily expandable to a 256-input /output network The chip uses on-chip SRAM for storing weights to reduce offdhip time delays and simultaneously takes advantage of a large and fast internal bus structure to move weights. The technology exploits InP for its high speed multipliers and multi chip module ( M C M ) packaging for interconnecting the ACUk Sigdicant progress has been ma& in the implementation of Artificial Neural Networks ( ANN ). However, the work reported to date is inclined more towards analog approaches [l-31. The next phase is likely to be the digital / hybrid solutions to ANN implementations. To this end, research at Clemson University has emphasized the implementation of large-scale ANN'S, where network size ( in terms of weights and neurons ) becomes a considerable challenge to VLSI technology. An excellent summary is found in [4,5]. The scaling of an ANN is especially significant in imaging applications, where the volume of image data leads to large ANNs [6].Rather than concentrate on a specific ANN architecture, the effort involves identification of the common computational characteristics of layered and recurrent networks and the expansion or scaling this computation to networks of arbitrary size. The solution strongly depend on the ability to compute inner products and then transfer these results to other units which do the same. Training of the neural units is assumed to occur off-chip. Other topics considered in the design phase include: weight and arithmetic accuracy, speed-area tradeoff and analog vs digital implementation. Some design issues are summarized in [7].