Pedestrian Recognition Integrating High-definition LIDAR and Vision

高解像度レーザレーダと画像の統合による歩行者認識

Pedestrian recognition is one of the key technologies for driving assistance systems. This paper proposes a pedestrian recognition system using a high-definition LIDAR and a vision sensor to achieve high performance in various conditions. Pedestrian candidates are extracted from two sensors in parallel by the SVM-based classifiers. In particular, the region-of-interests in the image processing is set by information about objects derived from LIDAR to reduce false positives as well as computational burden. All candidates are integrated by their likelihood calculated from their classification scores. A quantitative evaluation in a road environment confirms the effectiveness of the proposed system.高い認識性能を実現するため,レーザレーダと画像を統合した歩行者検出手法を提案する.提案手法は各センサから並行して歩行者候補を抽出する.このとき,レーダの立体物情報を用いて画像の誤検出を効率的に削減する.その後,候補に対する尤度に基づいて結果を統合する.走行環境での定量評価により提案手法の有効性を示す.

Language

  • English
  • Japanese

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01516209
  • Record Type: Publication
  • Source Agency: Japan Science and Technology Agency (JST)
  • Files: TRIS, JSTAGE
  • Created Date: Feb 28 2014 12:30PM