Skip to content
Skip to navigation menu

Physics Seminar

From galaxies to precision surfaces: 'Single-shot areal profilometry using hyperspectral interferometry with micro-optics arrays'

Speaker: Dr Pablo Ruiz, Dr Charles Russ Coggrave (Loughborough University)
Date: Wednesday 19 February 2020
Time: 15:00
Venue: N/3.23

I will describe a single-shot technique to measure areal profiles on optically smooth and stepped surfaces for applications where rapid data acquisition in non-cooperative environments is essential, e.g. for quality control in the production line, where vibrations impede traditional approaches. It is based on hyperspectral interferometry (HSI), a technique in which the output of a white-light interferometer provides the input to a hyperspectral imaging system. Previous HSI implementations suffered from inefficient utilisation of the available pixels which limited the number of measured coordinates and/or unambiguous depth range. Inspired by astronomy instrumentation (integral field units or imaging spectrographs) used to measure rotational velocity of galaxies, we developed a high throughput, single-shot system to measure surface height profile and local roughness without any sort of scanning required.

The second half of the talk will be given by Dr C.R.Coggrave with the title 'High-speed fringe projection and real-time phase analysis for robot 3D vision applications'

Vision systems based on projected structured light have become an established technique for 3D optical profilometry and 3D scene mapping. System integrators can access off-the-shelf solutions capable of providing dense megapixel point clouds updated several times per second with precision better than 100 micrometers. These systems have found many applications in quality control and non-destructive testing for high-value manufacturing. We have recently developed an algorithm based on maximum-likelihood analysis of the acquired 3D point clouds for automated recognition of objects, and estimation of their pose. However, adoption of these techniques in robotic applications within the manufacturing sector for scene analysis, assembly, path-planning, and surface measurement has been hindered by the rate at which independent 3D point clouds can be captured for the complex geometries that are typically associated with engineering components. Increasing the point cloud update rate is particularly attractive in dynamic workspaces such as moving conveyor systems and human-robot interaction. In this study, we are targeting a 3D vision system with an order of magnitude increase in throughput over the current state of the art. This has involved the development of diffractive optical elements to achieve sinusoidal fringe pattern projection rates of up to 20 kHz, image acquisition at framing rates of over 3kHz, and real-time implementation of the underlying temporal phase unwrapping algorithm on a GPU to deliver high-precision independent megapixel 3D point clouds at >90Hz, representing a coordinate acquisition rate of over 10^8 s^-1.