Mask R-CNN Based Real Time near Drowning Person Detection System in Swimming Pools

Abstract

To find the drowning person in time in swimming pool to reduce the drowning person mortality rate. We used the Mask R-CNN algorithm, and optimizing the convolution backbone of the traditional Mask R-CNN algorithm by adding features of cascaded with pyramid model to design a swimmer drowning detection system. Through real-time recognition of the posture of swimmers in the swimming pool, it can determine the drowning person and alert in time. The system's proposed algorithm has been put to the test on multiple real-world video sequences taken in swimming pools, and the findings show that it is very accurate and capable of monitoring people in real time. The experimental results show that the detection speed of the system is 6 FPS, while the detection rate is 94.1 %, while the false detection rate is 5.9%. The effect is good, which satisfying the anticipated requirements.