Real-time pedestrian warning system on highway using deep learning methods

Abstract

To lower the traffic accidents in highway systems, it is important to assure the highway be used only by vehicles. If someone accidentally enters the highway without noticing the potential danger, some traffic management system may give out an alarm to the pedestrian or to the nearby vehicles. That can be achieved by modern technology. That is, if the monitoring system or car camera can capture the pedestrian information and immediate give an alarm, obviously it can effectively reduce the incidence of accidents. For this purpose, in this paper, we propose a pedestrian detection algorithm with optimized detection method of region-convolution neural network. It is demonstrated by experiments that the proposed method is able to reach the state-of-the-art methods level. Finally, we implement this algorithm to a real-time monitoring system that could realize pedestrian saliency detection and alarm immediately on the entrance, exits and other important places ofhighway.