• DocumentCode
    1844631
  • Title

    A method of location the vehicle windshield region for vehicle occupant detection system

  • Author

    Xue Yuan ; Yifei Meng ; Xueye Wei

  • Author_Institution
    Sch. of Electron. & Inf. Eng., Beijing Jiaotong Univ., Beijing, China
  • Volume
    1
  • fYear
    2012
  • fDate
    21-25 Oct. 2012
  • Firstpage
    712
  • Lastpage
    715
  • Abstract
    This paper investigates a maximum energy method to extract the windshield regions from the vehicles, proposed method combines the information of color (gray-scale), shape and complexity for location the vehicle windshield region, which can be applied vehicle occupant detection and recognition system successfully. Firstly, the mean-shift segmentation method is adopted to divide the vehicle image into several segments, and Hough transform method is adopted to extract the shape lines from edges of vehicle, these shape lines compose several trapezoids as candidate regions. Then, the maximum energy method is proposed to select final windshield region from the candidate regions. Finally, Histograms of Oriented Gradients (HOG) descriptors based method is adopted to detect the occupants in the extracted windshield regions.
  • Keywords
    Hough transforms; feature extraction; gradient methods; image classification; image segmentation; object detection; object recognition; shape recognition; traffic engineering computing; HOG descriptor based method; Hough transform method; color information; complexity information; edge detection; histogram of oriented gradient; maximum energy method; mean shift segmentation method; shape information; shape line extraction; vehicle image division; vehicle occupant detection system; vehicle recognition system; vehicle windshield region location; Hough transform; maximum energy method; mean-shift segmentation; vehicle occupant detection; vehicle windshield detection;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Signal Processing (ICSP), 2012 IEEE 11th International Conference on
  • Conference_Location
    Beijing
  • ISSN
    2164-5221
  • Print_ISBN
    978-1-4673-2196-9
  • Type

    conf

  • DOI
    10.1109/ICoSP.2012.6491586
  • Filename
    6491586