Gnarometer Surfcam Live Inspector

This paper proposes a real-time method for wave detection and tracking from pre-existing operational surf cameras. Firstly, an adaptive Gaussian mixture model with background history and morphological opening is used to subtract the background from the raw frame. Contour finding is applied to the white foreground pixels to identify objects and area and inertia filtering is used to remove false positive contours. The contours which remain are converted to wave objects using a wave attribute class. The centre of mass and the original angle of the wave's axis is used to reduce the search area for the wave's location for each successive frame to a singular dimension along an estimated axis. Displacement, white pixel mass, length and time are compared against set user thresholds to verify the tracked wave is a true instance of a wave. The characteristics of the recognised waves are displayed on the frame and exported to an external text document. The results show that the method detects 90% of waves on high quality cameras, and 61.5% on low quality and long distance cameras, exceeding previous research conducted by P. Green which detected a maximum of 55% of observed waves. Performance of the proposed method was therefore more accurate and extensive than related works, calculating a wider variation of wave attributes without requiring any additional hardware.