An Experimental Study on Pedestrian Classification

This article presents an in-depth experimental study on detecting people in images, resulting in pedestrian classification. The authors examined multiple feature-classifier combinations with respect to their ROC performance and efficiency. The authors focus on global versus local and adaptive versus nonadaptive features, as exemplified by principal component analysis (PCA) coefficients, Haar wavelets, and local receptive fields (LRFs). These experiments were performed on a large data set consisting of 4,000 pedestrian and more than 25,000 nonpedestrian (labeled) images capture in outdoor urban environments. The authors note that the ability to detect people in images is key to a number of important applications including surveillance, robotics, and intelligent vehicle design. The authors conclude that the combination of Support Vector Machines (SVMs) with LRF features creates the best results.


  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01047269
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Apr 27 2007 2:20PM