
Abstract
Low-cost camera systems based on off-the-shelf imaging components are seeing increasingly widespread use for metrology orientated research purposes such as robotics, part alignment and inspection. This presentation describes a system under continuing development at University College London. It gives first results from laboratory testing of a constellation of four synchronised off-the-shelf cameras connected to a $1000 computer to image at 100Hz. This constellation has the goal of achieving on-line sub-mm measurement of multiple, rapidly moving and deforming test objects.
The exploratory test setup comprises four 3MP USB3.0 cameras interfaced independently at 5Gbs to a multiport USB3.1 card for processing on an i7 intel processor. Cameras are synchronised together with a custom synchronisation box based on an LED lighting controller sending a timing trigger pulse in parallel to all cameras. Illumination is based on low-cost $10 green LED ring lights mounted using 3D prints to the front of each camera lens. The system is used to observe two $20 USB-powered 4” desk fans which independently rotate in the field of view of the cameras. Blocks of coded targets are located at a 2” radius on the fan blades. When rotating at around 2700rpm these deliver a convenient velocity of 13ms-1 with each fan running unsynchronised and independently.
Cameras are calibrated achieving an RMS image coordinate precision of 0.17mm with a conventional photogrammetric self-calibrating bundle adjustment taking input measurements from a 3D targeted reference object moved around the working volume. The object then provides a convenient static reference against which the cameras are automatically orientated. The two fans are positioned arbitrarily, close to the reference object and within the field of view of all four cameras. This gives four intersecting lines of sight per target from which the pose of the cameras, the 3D target coordinates on the moving fan blades and the 6DoF of each fan rotation are computed (see Figure 1). The objective is to compute this for every incoming set of images at 100Hz. Currently the system is limited by the 20ms computation time needed to robustly detect, identify and measure 2D target image locations across all four images and to display and update the user interface. Subsequent 3D processing, including 3D display, takes 6ms. This combination gives an output data frequency of 35Hz with a multithreaded i7 CPU load of 25 %. Work is ongoing to further parallelise the 2D processing pipeline, making use of available CPU capacity and drawing on GPU where necessary, so that the output frequency can match the input frame rate.
In the presented example, output 6DoF data is transmitted through Hexagon/NRK’s pipeline instrument interface to Spatial Analyser Software at 35Hz. This test shows a significant difference between the rotation-induced vibration (and noise level) of the two fans being tested (see Figure 2).
This project (20IND02 – DynaMITE) has received funding from the EMPIR programme co-financed by the Participating States and from the European Union's Horizon 2020 research and innovation programme. This project is a collaboration across NMIs and RTOs to expand the capability of optical metrology systems with a view to future industrial applications supported by the UK’s Royal Academy of Engineering and Airbus UK which together sponsor the UCL Research Chair in Large-Volume Metrology.
Figure 1: Synchronized 102Hz photogrammetric views of rotating fan blades and static reference targets with in-line 3D view. Estimated 6DoF RMS based on tracking 32 targets on the rotating blades and comparing with reference values is 0.02mm
Figure 2: 6DoF frames representing fan rotation centers transmitted to a remote computer running NRK Spatial Analyzer software. These data comprise 3,600 6DoF frames at 35Hz. Fan 1 [left: max translation 0.17mm] exhibits vibration and increased noise over fan 2 [right: max translation 0.1mm] which runs quietly and describes a smoother rotation.
Presentation media
3 video clips from this presentation are available to CMS members.
Cite this paper
Stuart Robson, Ben Sargeant, R. Smith, Stephen Kyle, Mark Shortis. University College London, UK; RMIT University, Melbourne, Australia. Presented at the Coordinate Metrology Systems Conference 2022, Orlando, Florida, July 25-29, 2022. Coordinate Metrology Society, reference CMS-2022-013.
