Search Results

In this work, we outline a process for traffic light detection in the context of autonomous vehicles and driver assistance technology features. For our approach, we leverage the automatic annotations from virtually generated data of road scenes. Using the automatically generated bounding boxes around the illuminated traffic lights themselves, we trained an 8-layer deep neural network, without pre-training, for classification of traffic light signals (green, amber, red). After training on virtual data, we tested the network on real world data collected from a forward facing camera on a vehicle. Our new region proposal technique uses color space conversion and contour extraction to identify candidate regions to feed to the deep neural network classifier. Depending on time of day, we convert our RGB images in order to more accurately extract the appropriate regions of interest and filter them based on color, shape and size. These candidate regions are fed to a deep neural network.

Increased awareness of health effects caused by airborne contaminants that include natural and industrial aerosols, bioaerosols and gases, has led to increased usage of various kinds of filters. This trend is reflected in the automotive industry, where cabin air filters are increasingly offered as a means to reduce the likelihood of inhaling these contaminants while driving. Pleated filters, typically employing charge enhanced, thermoplastic base non woven media, have most commonly been applied in order to achieve the requisite level of particle capture, at minimum expense of additional burden to the vehicle HVAC system. The reliability of these filters, however, has been under scrutiny. This is particularly true for those derived from depth electrostatic media. In this study we have evaluated a newly developed depth media, as well as a split fiber electret media, under various simulated environmental and loading conditions.