The representation, recognition, and locating of 3-d objects
International Journal of Robotics Research
Three-dimensional object recognition from single two-dimensional images
Artificial Intelligence
New methods for matching 3-D objects with single perspective views
IEEE Transactions on Pattern Analysis and Machine Intelligence
Pose Determination of a Three-Dimensional Object Using Triangle Pairs
IEEE Transactions on Pattern Analysis and Machine Intelligence
An analytic solution for the perspective 4-point problem
Computer Vision, Graphics, and Image Processing
Determination of the Attitude of 3D Objects from a Single Perspective View
IEEE Transactions on Pattern Analysis and Machine Intelligence
Rambo-vision and planning on the connection machine
Proceedings of a workshop on Image understanding workshop
Exact and Approximate Solutions of the Perspective-Three-Point Problem
IEEE Transactions on Pattern Analysis and Machine Intelligence
Some necessary conditions on the number of solutions for the P4P problem
IWMM'04/GIAE'04 Proceedings of the 6th international conference on Computer Algebra and Geometric Algebra with Applications
Hi-index | 0.10 |
This paper describes a selective generate-and-test algorithm for SIMD-parallel calculation of the 3-D pose of a known object from a single perspective view. The algorithm consists of three main stages: object pose estimation, optimization, and reliability analysis. The first stage involves parallel generate-and-test of candidate pose solutions obtained by selectively matching model and image point triples, testing their correspondence by parallel transformation of all visible model points and comparison of their features with the image. Instead of exact algebraic calculations, the three point perspective pose estimation problem is solved numerically. In the second stage, the initial pose estimate is optimized using a least-squares method. In the final stage, the reliability of the optimized pose is estimated using covariance analysis. The simulations showed that the approximate initial pose estimates are sufficiently good to obtain very accurate optimized results. Furthermore, the processing times for an object with 12 vertices were on the order of a few hundreds of milliseconds on a Connection Machine-2 with 512 floating point units.