Design of torrent detection and alarm device based on deep learning

In order to enhance the safety of personnel involved in water rescue and training missions, it is necessary to rapidly detect and forecast hazardous waters in real time. In this study, a device for detecting and alerting to torrents has been developed using a deep learning object detection algorithm. Through target detection, dangerous points in torrent waters are identified and recorded. Furthermore, the intensity of rapids is analyzed in order to enhance perceptual ability. The findings indicate that the convolutional neural network is capable of successfully detecting and recognizing discontinuous target torrents. In field measurements, the device demonstrates a strong sensing alarm effect. The practical significance lies in further enhancing the rescue personnel’s ability to perceive and predict the risks of hazardous factors such as torrent. The application of the device also improves the efficiency and safety of water rescue.