B David Prakash, IAG Firemark, Singapore

Safety is a critical aspect in all swimming pools. This paper describes a near-drowning early prediction technique using novel equations (NEPTUNE). NEPTUNE uses equations or rules that would be able to detect near-drowning using at least 1 but not more than 5 seconds of video sequence with no false positives. The backbone of NEPTUNE encompasses a mix of statistical image processing to merge images for a video sequence followed by K-means clustering to extract segments in the merged image and finally a revisit to statistical image processing to derive variables for every segment. These variables would be used by the equations to identify near-drowning. NEPTUNE has the potential to be integrated into a swimming pool camera system that would send an alarm to the lifeguards for early response so that the likelihood of recovery is high.

Near-drowning Detection, Drowning Detection, Statistical Image Processing, K-means Clustering, Swimming Pools