In the early 1980's, with the advent of high speed Very Large Scale Integrated Circuits, real-time image processing and pattern recognition became possible. These techniques can be used to significantly improve the performance of movement analysis equipment by increasing the reliability of marker detection. It has been shown [1] that the use of the cross-correlation algorithm for shape recognition improves the accuracy of marker coordinates. Furthermore, when shape recognition is used, smaller targets can be tracked without degradations in performance. The ELITE system employs a two-level, hierarchical, structure to recover 3D coordinates from images. The first level of this structure is contained in the system hardware which processes the data using a Single Instruction Multiple Data approach. This permits an unlimited number of markers to be detected in each frame, at up to 100 frames per second (Hz). The second level of the structure is contained in the system software installed in a general purpose personal computer (PC). This level corrects image distortions; establishes marker identities; performs the required tracking, space resection and intersection; filters the data stream; and provides graphic displays of results. The overall accuracy of the 3D system is estimated to be 1 part in 2800 over the field of view [2].
|