3D MOTION ESTIMATION OF A MOVING OBJECT USING MODEL-BASED TECHNIQUES

This paper presents a method of three-dimensional (3D) motion estimation of a moving object using model-based image coding. This is based on localization of a viewpoint from monocular vision. First, the whole scheme of the environment recognition system is described. This is composed of four major modules, including moving object detection, model-based image estimation, 3D model database, and space quantization. Second, the moving object detection module using block matching method is explained, which incorporates a two-dimensional motion compensation technique in video compression standard. Then, model-based image estimation is described. This is based on model-based image coding, which is considered the most powerful method to achieve the very low bit-rate image coding and transmission. This module includes texture mapping techniques using 3D model database and source texture images. Next, the space quantization method is modified to avoid difficulty in searching six parameters when the target object moves widely in outdoor environments. This modification is performed by incorporating virtual-real-images into the image estimation process, and by modification of the search process as well as error evaluation function. Finally, an example of a moving car demonstrates the potential of 3D motion estimation method for environment recognition.

  • Supplemental Notes:
    • Five volumes of papers and one volume of abstracts comprise the published set of conference materials.
  • Corporate Authors:

    VERTIS

    TORANOMOM 34 MORI BUILDING 1-25-5
    TORANOMON, MINATOKU, TOKYO 105  Japan 
  • Authors:
    • SASAKI, M
    • Toyoda, A
  • Conference:
  • Publication Date: 1995-11

Language

  • English

Media Info

  • Pagination: p. 1075

Subject/Index Terms

Filing Info

  • Accession Number: 00722041
  • Record Type: Publication
  • Report/Paper Numbers: Volume 3
  • Files: TRIS
  • Created Date: Jun 25 2000 12:00AM