A FULLYAUTOMATED MOTION ANALYSIS SYSTEM FOR HAND HELD CAMERAS

  • John Newton
  • Ricky Turner

Abstract

Traditionally much of the biomechanical study of human movement has relied on collecting film or video data using very stable cameras mounted on tripods. The geometrical positioning of the camera(s) and the introduction of complex reference frames have been to justify the accuracy of motion data. The data is then subjected to an' often time consuming digitisation process of questionable reliability. Automated tracking systems have almost entirely relied on specialist cameras with some form of sensor attachment to the moving object, restricting the system mainly to a laboratory/clinical environment. The inaccuracy and time consuming nature of manual digitisation systems has long been a concern of the biomechanist. An automated motion analysis system based on pixel recognition, using normal video systems, has been demonstrated by Turner and Newton (1992) and Newton, Turner and Greenwood (1994). None of these systems have addressed the problem of "Iarge volume space". Cameras are positioned in order to view the filming area without employing camera motion. This offen results in very small image data. A pan and tilt camera analysis system was used at the Calgary Winter Olympics (Yeadon 1989). More recently, systems which require cameras to be mounted on tripods with special "pan and tilt" heads have become commercially available. The new system, which is the subject of the presentation, is an extension of the method demonstrated by Turner and Newton (1992) and now allows for the automated analysis of motion sequences with NO constraints on camera position. Fixed cameras may pan, tilt and zoom, however, the camera does not need to be fixed and may be mobile or hand held. Motion sequences are automatically created from the video data alone without the use of external camera calibration. Once processed, measurements may be made directly off the motion data for 20 or 3D forms of analysis and presentation. The system also retains the desirable characteristic that it does not need any form of surface markers (reflective or otherwise) and does not require special cameras or Iighting. It will run on standard P. C's., workstations or portables and connect to any standard video source. It is also possible, within limits, to analyse previously recorded data such as commercial video, without previous knowledge of camera position. The new system has completed the initial stages of development and will be available for demonstration, and technical explanation, if given the opportunity, at ISBS'96. REFERENCES TURNER, R. and NEWTON, J. (1992). Automated 40 Motion Analysis Biomechanics in Gymnasts. Koln, Germany. NEWTON, J. TURNER, R. and GREENWOOD, M. (1994) An Automated 3D Motion Analysis System. Second World Congress of Biomechanics, Amsterdam YEADON, M. (1989). A Method for Obtaining Three -Dimensional Data on Ski -Jumping Using Pan and Tilt Cameras. IJSB 5:248 -247
Section
Equipment / Instrumentation