Combining Shape and Color Information For 3D Object Recognition

Both photometric and geometric information are important for 3D object recognition. Traditionally, however, few systems utilized both types of information. This is because no single representation is suitable for both types of information. This paper proposes a method for representing both color and geometric information using a common framework, the Spherical Attribute Image (SAI). The SAI maps the values of curvature and color computed at every node of a mesh approximating the object surface onto a spherical image. A model object and an observed surface are computed by finding the rotation that brings their spherical images into correspondence. We show how this matching algorithm can be used for object recognition using both geometric and photometric information. In addition, we describe how the two types of information can be combined in a way that takes into account their actual distribution on the surface.

[1]  G. Champleboux,et al.  From accurate range imaging sensor calibration to accurate model-based 3D object localization , 1992, Proceedings 1992 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[2]  Jake K. Aggarwal,et al.  TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE , 2008 .

[3]  K. Sato,et al.  Range imaging system utilizing nematic liquid crystal mask , 1987 .

[4]  Azriel Rosenfeld,et al.  Registration of multiple overlapping range images: scenes without distinctive features , 1989, Proceedings CVPR '89: IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[5]  Katsushi Ikeuchi,et al.  Shape representation and image segmentation using deformable surfaces , 1992, Image Vis. Comput..

[6]  Katsushi Ikeuchi,et al.  A spherical representation for the recognition of curved objects , 1993, 1993 (4th) International Conference on Computer Vision.

[7]  Katsushi Ikeuchi,et al.  Determining 3-D object pose using the complex extended Gaussian image , 1991, Proceedings. 1991 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[8]  Bahram Parvin,et al.  B-rep from unregistered multiple range images , 1992, Proceedings 1992 IEEE International Conference on Robotics and Automation.

[9]  Azriel Rosenfeld,et al.  Registration of Multiple Overlapping Range Images: Scenes Without Distinctive Features , 1991, IEEE Trans. Pattern Anal. Mach. Intell..

[10]  Katsushi Ikeuchi.  Recognition of 3-D Objects Using the Extended Gaussian Image , 1981, IJCAI.

[11]  Paul J. Besl,et al.  A Method for Registration of 3-D Shapes , 1992, IEEE Trans. Pattern Anal. Mach. Intell..

[12]  Gérard G. Medioni,et al.  Object modelling by registration of multiple range images , 1992, Image Vis. Comput..