Development of object detection model for sizing safety measures in Human-Industrial mobile robot interaction

In human -robot interaction, single -level safety measures are traditionally applied, and employee -specific criteria are not taken into account. However, a new method can be developed using object detection technology, and the risk level of human -robot interaction can be determined by identifying employeespecific criteria such as the use of protective equipment and authorization levels, and different -sized safety measures can be applied depending on the risk magnitude. In this study, YOLOv5n, YOLOv8n, and SSD MobileNet V3 object detection models were developed and analyzed for this purpose. The results show that architectures belonging to the YOLO family run faster and achieve higher levels of accuracy. The YOLOv5n algorithm achieved a speed of 650 FPS with the use of a GPU and an F1 accuracy of 95.7% as a result of the evaluation with test data. The results show that object detection technology has reached an accuracy and speed that can be applied simultaneously with proximity senors, and that industrial mobile robots can detect worker characteristics and rate risks before taking safety measures. This allows for safer working environments, eliminates unnecessary precautions, and optimizes operational efficiency. In addition, this method can be applied in many sectors and areas to provide safe working environments.