2018/08/27 by Mizanur Rahman, Mhafuzul Islam, Rahman, Mizanur +5
Computer Science · Engineering · #Advanced Neural Network Applications #Autonomous Vehicle Technology and Safety #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Fire Detection and Safety Systems #Video Surveillance and Tracking Methods
paper · pdf · doi:10.48550/arxiv.1808.09023
openalex publication_date 2018/08/27 · openalex created_date 2022/08/03 · openalex updated_date 2026/07/28
Vehicle-to-Pedestrian (V2P) communication can significantly improve\npedestrian safety at a signalized intersection. It is unlikely that pedestrians\nwill carry a low latency communication enabled device and activate a pedestrian\nsafety application in their hand-held device all the time. Because of this\nlimitation, multiple traffic cameras at the signalized intersection can be used\nto accurately detect and locate pedestrians using deep learning and broadcast\nsafety alerts related to pedestrians to warn connected and automated vehicles\naround a signalized intersection. However, unavailability of high-performance\ncomputing infrastructure at the roadside and limited network bandwidth between\ntraffic cameras and the computing infrastructure limits the ability of\nreal-time data streaming and processing for pedestrian detection. In this\npaper, we develop an edge computing based real-time pedestrian detection\nstrategy combining pedestrian detection algorithm using deep learning and an\nefficient data communication approach to reduce bandwidth requirements while\nmaintaining a high object detection accuracy. We utilize a lossy compression\ntechnique on traffic camera data to determine the tradeoff between the\nreduction of the communication bandwidth requirements and a defined object\ndetection accuracy. The performance of the pedestrian-detection strategy is\nmeasured in terms of pedestrian classification accuracy with varying peak\nsignal-to-noise ratios. The analyses reveal that we detect pedestrians by\nmaintaining a defined detection accuracy with a peak signal-to-noise ratio\n(PSNR) 43 dB while reducing the communication bandwidth from 9.82 Mbits/sec to\n0.31 Mbits/sec, a 31x reduction.\n