Safety Detection Using Convolutional Neural Network and Dual-Polarized Miniaturized Antennas Integrated in Metal Carabiner

Safety detection is very important for high-altitude operations due to the high risk of falling. Workers are required to attach the carabiner of a safety belt to a metal structure. Although sensors are added to the carabiner to detect whether it is attached to an object, workers may attach the carabiner to their arm to cheat the sensors for convenience. This paper presents a more accurate detection method using a convolutional neural network (CNN) to analyze the scattering parameters of dual-polarized antennas. A dual-polarized substrate integrated waveguide (SIW) antenna is designed and integrated into a metal carabiner. The CNN is designed to analyze the scattering parameters, utilizing multi-scale convolution and attention mechanisms to classify safe or unsafe situations with high accuracy. Experimental results demonstrate classification accuracies of 96.2% on simulated datasets and 90% in measurement. The proposed method has advantages in non-contact detection and metal integration.