Design of drowning detection method based on YOLOv8

This article introduces a real-time drowning detection system based on the YOLOv8-POSE model and OpenCV. The system analyzes key points of human body posture, monitors swimmers’ posture in real time, and detects abnormal postures to identify potential drowning events. Afterwards, it will be handed over to a specially trained drowning recognition model to make targeted judgments on whether drowning has occurred. The system design goal is to improve the safety monitoring capability of water bodies such as swimming pools and beaches, by alerting lifeguards or monitoring systems through visual warnings and alarms. To solve the problems of incomplete coverage, slow response time, high labor intensity, false alarms and omissions in traditional manual drowning prevention inspections, as well as the limitation of time and location in manual supervision are proposed, in order to improve the response speed and rescue success rate of drowning incidents.