Date of Award

8-2007

Document Type

Thesis

Degree Name

Master of Science (MS)

Legacy Department

Mechanical Engineering

Advisor

Kurfess, Thomas R

Committee Member

Tucker , Thomas M

Committee Member

Mocko , Gregory M

Abstract

Data point set registration is an important operation in coordinate metrology. Registration is the operation by which sampled point clouds are aligned with a CAD model by a 4X4 homogeneous transformation (e.g., rotation and translation). This alignment permits validation of the produced artifact's geometry. State-of-the-art metrology systems are now capable of generating thousands, if not millions, of data points during an inspection operation, resulting in increased computational power to fully utilize these larger data sets. The registration process is an iterative nonlinear optimization operation having an execution time directly related to the number of points processed and CAD model complexity. The objective function to be minimized by this optimization is the sum of the square distances between each point in the point cloud and the closest surface in the CAD model. A brute force approach to registration, which is often used, is to compute the minimum distance between each point and each surface in the CAD model. As point cloud sizes and CAD model complexity increase, this approach becomes intractable and inefficient. Highly efficient numerical and analytical gradient based algorithms exist and their goal is to convergence to an optimal solution in minimum time.
This thesis presents a new approach to efficiently perform the registration process by employing readily available computer hardware, the graphical processor unit (GPU). The data point set registration time for the GPU shows a significant improvement (around 15-20 times) over typical CPU performance. Efficient GPU programming decreases the complexity of the steps and improves the rate of convergence of the existing algorithms. The experimental setup reveals the exponential increasing nature of the CPU and the linear performance of the GPU in various aspects of an algorithm. The importance of CPU in the GPU programming is highlighted.
The future implementations disclose the possible extensions of a GPU for higher order and complex coordinate metrology algorithms.

Share

COinS