Fast Pedestrian Detection for Mobile Devices

In this paper the authors present a fast and robust solution for pedestrian detection that can run in real time conditions even on mobile devices with limited computational power. An optimization of the channel features based multiscale detection schemes is proposed by using 8 detection models for each half octave scales. The image features have to be computed only once each half octave and there is no need for feature approximation. The authors use multiscale square features for training the multiresolution pedestrian classifiers. The proposed solution achieves state of art detection results on Caltech pedestrian benchmark at over 100 FPS using a CPU implementation, being the fastest detection approach on the benchmark. The solution is fast enough to perform under real time conditions on mobile platforms, yet preserving its robustness. The full detection process can run at over 20 FPS on a quad-core ARM CPU based smartphone or tablet, being a suitable solution for limited computational power mobile devices or embedded platforms.

Language

  • English

Media Info

  • Media Type: Web
  • Features: References;
  • Pagination: pp 2364-2369
  • Monograph Title: 18th International IEEE Conference on Intelligent Transportation Systems (ITSC 2015)

Subject/Index Terms

Filing Info

  • Accession Number: 01602804
  • Record Type: Publication
  • ISBN: 9781467365956
  • Files: TRIS
  • Created Date: May 2 2016 3:18PM