Bow movement while aiming, three methods: Ganter 2010

Three archers from the German women's recurve national squad (B squad) shot 12 arrows each at 30 metres indoors while three different systems recorded the bow's movement during aiming at the same time: a Noptel optoelectronic device mounted on the bow, a 3D video analysis with two cameras and a 2D video with a single camera. For each shot the authors calculated how much the aim moved in the last second before release, horizontally and vertically, and checked how well the three systems agreed, archer by archer.

Noptel and 3D video agree well: correlations for horizontal movement are 0.80, 0.51 and 0.92, and for vertical movement 0.60, 0.91 and 0.69. The 2D video follows the Noptel vertically (0.69, 0.97 and 0.89) and the 3D less evenly (0.18, 0.94 and 0.54), but horizontally it follows neither: correlations range from -0.31 to 0.29. The archer with the poorest horizontal agreement was also the one who, in the 3D data, rotated the bow most in the frontal plane. The three archers' mean scores were 9.33, 9.5 and 9.17.

The modified Noptel ST-2000 weighs about 160 grams and fits in place of the lower stabiliser: an infrared beam strikes prisms near the target and returns the aiming trace on the target at 66.6 Hz for the 4 seconds before release, which the device detects from the bow's vibration with a 15 ms delay. For the 3D video, two Sony cameras at 25 frames per second (analysed as 50 half-frames), one diagonally behind the archer and one to the side, 5 to 6 metres away at full zoom, tracked four 3 mm markers fixed to the sight bar; the estimated error is half a millimetre at most. The 2D video used only the diagonal camera: the sight was located in each frame with a template-matching algorithm in LabVIEW and its displacement multiplied by 30 to project it onto the target.

The target was the 60 cm FITA face, with a 6 cm ten ring. The measures compared were the standard deviations of horizontal and vertical movement in the last second and the correlations between the aiming time series, all calculated separately for each archer over her 12 shots.

Three archers, 12 arrows each, a single indoor distance: the findings describe how well the systems agree under these conditions, not in competition or at 70 metres. Each correlation rests on 12 shots and so carries a wide margin of uncertainty. None of the three systems serves as an absolute reference: the Noptel is treated as the benchmark, but it adds weight to the bow and does not separate translation from rotation, so its agreement with the 3D system shows that two methods give similar results, not that either measures aiming without error.

The explanation for the 2D system's horizontal discrepancies, rotations of the bow about the sight that a single camera cannot see, is plausible but rests on one archer. The study does not relate aiming movement to score.

For archers who film themselves to check their aim, the finding is concrete: video from a single camera, here placed diagonally behind the archer, captures reasonably well how much the sight moves up and down in the final moments, but not how much it moves sideways, because the bow can rotate and the camera sees only displacement. A sight that looks steady horizontally on video may therefore go with an aim that is moving on the target.

For coaches, the study shows that the Noptel and 3D analysis give a similar picture of aiming movement. The Noptel, however, adds weight to the bow and cannot be used in competition, while 3D takes a long time to set up and process and gives no immediate feedback; 2D video is the only one of the three usable in competition, with its horizontal limitation.