Parallel algorithm for object recognition and its implementation on a MIMD machine

The PERFORM matching method, introduced by B. Modayur and I.G. Shapiro (1994), solves the recognition problem under a bounded error noise model by establishing correspondences between model and image features. PERFORM evaluates correspondences by intersecting error regions in the image space. The article describes the parallel formulation of the PERFORM matching method and the implications of a shared memory, MIMD implementation. When a single solution is sought, the time complexity of the sequential matching algorithm using point features is of the order O(I/sup 2/NI) for 2D-2D matching and O(I/sup 3/NI) for 2D-3D matching, where N is the number of model features and I is the number of image features. The corresponding parallel algorithm using O(I/sup 2/) processors for 2D-2D matching and processors for 2D-3D matching has O(NI) complexity. When line features are used, the sequential complexity is of the order O(I NI) for 2D-2D matching and O(I/sup 2/ NI) for 2D-3D matching. The corresponding parallel algorithm utilizing O(I) processors for 2D-2D matching and O(I/sup 2/) processors for 2D-3D matching has O(NI) complexity. When implemented in parallel, the method requires minimal memory and obviates load balancing overheads and communication between processors. The article describes parallel implementations of 2D-2D matching on a shared memory, MIMD machine (KSR-I). Results show that significant, close to linear speedups are achievable using multiple processors.

[1]  Clark F. Olson.  Time and space efficient pose clustering , 1994, 1994 Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.

[2]  David W. Jacobs,et al.  Error propagation in full 3D-from-2D object recognition , 1994, 1994 Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.

[3]  W. Eric L. Grimson,et al.  A Study of Affine Matching With Bounded Sensor Error , 1992, ECCV.

[4]  Thomas M. Breuel,et al.  Fast recognition using adaptive subdivisions of transformation space , 1992, Proceedings 1992 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[5]  Robert A. Hummel,et al.  Massively parallel model matching: geometric hashing on the Connection Machine , 1992, Computer.

[6]  H. Baird.  Model-Based Image Matching Using Location , 1985 .

[7]  Tomás Lozano-Pérez,et al.  Tactile Recognition and Localization Using Object Models: The Case of Polyhedra on a Plane , 1984, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[8]  W. Eric L. Grimson,et al.  Localizing Overlapping Parts by Searching the Interpretation Tree , 1987, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[9]  Linda G. Shapiro,et al.  Fast parallel object recognition , 1994, Proceedings of the 12th IAPR International Conference on Pattern Recognition, Vol. 2 - Conference B: Computer Vision & Image Processing. (Cat. No.94CH3440-5).

[10]  L. W. Tucker,et al.  Object recognition using the Connection Machine , 1988, Proceedings CVPR '88: The Computer Society Conference on Computer Vision and Pattern Recognition.