Accuracy of a portable (PTZ digital) camera system designed for aquatic three-dimensional analysis

Three-dimensional (3D) motion analysis of aquatic activities such as swimming requires high accuracy throughout large volumes particularly when above and below water data are merged for full body analysis. The purpose of this study was to assess the accuracy of a recently developed portable camera system designed for 3D kinematic data collection of aquatic activities including swimming. The system comprises of 6 (Elmo PTC-450C, Elmo CO., Ltd, Nagoya, Japan) cameras with 12x optical zoom lenses (f=3.8-45.6mm, F=1.6-2.7). For 3D reconstruction for analysis of swimming a 14 segment body model is digitized for two above and four below water cameras. The underwater cameras are housed in perspex boxes. Given that distortion of the perspex box can increase the errors in calibrating the 3D space it is essential that errors in 3D reconstruction of the points using the standard direct linear transformation (DLT) approach are known. For both above and below water views twelve markers with known 3D coordinate locations in a 6.75 cm3 calibration frame were digitized and the data input to a DLT programme. A different set of twelve markers with known locations were then digitized for both above and below water views and their 3D locations calculated using the DLT equations generated from the digitized data of the original set of markers. The root mean square errors (RMS) errors in estimating the locations of the second set of markers relative to their known locations were then quantified. Concurrently the same procedure was conducted for data recorded by a second camera system for which high accuracy and reliability has been established previously (Psycharakis, Sanders and Mill, 2005). Conclusions regarding the accuracy of the new system were drawn based on the magnitude of the RMS errors with reference to the accuracy of the established system.