VERPOSE prototype

NEWS:
Our ART colleague Dr. Steffen Herbort will give a lecture about "Markerless Tracking and intelligent screw driving assistance for industry 4.0 scenarios" at the FAPS symposium in Nürnberg on July 26 & 27!

 

VERPOSE (verify position) prototype is a tool for identifying operating points in a manufacturing process, for quality assurance and documentation.  It was developed as part of the ARVIDA project.

The VERPOSE camera, attached to the worker´s tool,  observes the vicinity  of the actual working point and compares the image with trained data. Verpose answers the question: On which of the (trained) working points is the tool actually located?

ART developed both the Verpose camera and the Verpose classification software. Depending on the image characteristics of the set of operating points, suitable matching algorithms are applied which results in a robust real time classification.


Performance

  • > 99,5% recognition rate
    (evaluated by automotive industry end-users)
  • robust detection method
    • any orientation of the tool
    • unaffected by ambient lighting
    • can be used with any colour of painted surface
    • stable if artifacts occur
  • flexible and adaptable for many different industrial tool-tracking applications 

Requirements

Components for VERPOSE:

  • VERPOSE camera installed on the respective tool
  • VERPOSE computer (industrial PC)
  • VERPOSE software

VERPOSE camera
  • integrated IR illumination
  • robust aluminium housing
  • adapter for a variety of EC screw drivers
  • integrated inertial sensor
  • cable length max. 100m
VERPOSE computer
  • industrial specification PC
VERPOSE software Training
  • (service of ART)
  • recording of the training data set
  • evaluation & pre-processing of the training images
  • training of the classifiers
Operation
  • evaluation of the camera images
    • in realtime: > 15 HZ
    • recognition rate: > 99,5%
    • actually for max. ~10 operating points
  • verification by comparison with classifiers of previous frames
  • possible extension of the training image records