University Archives

Safe collaboration between humans and machines remains a central challenge across industrial, healthcare, and domestic environments. From workshop tools to service robots, humans continue to face risks when working near moving or rotating mechanisms. Addressing these hazards requires intelligent systems capable of perceiving human presence, understanding object geometry, and enforcing safety boundaries in real time. This work presents an AI-driven safety framework that unites computer vision, depth sensing, and intelligent control to enable safe coexistence between humans and machines. Two complementary applications are demonstrated: a drill press safety system and a robotic tool handover system. Both utilize an Intel RealSense D415 depth camera and a custom-trained YOLO (You Only Look Once) segmentation model to detect and localize human hands, tools, and machine components. In the drill press implementation, it verifies the presence of hold-down clamps before operation and automatically disables power through a relay interface whenever a hand enters the defined danger zone. In the robotic handover implementation, the YOLO model segments each object (knives and screwdrivers are used in the demonstration system) into safe, pick, and sharp zones. The robot interprets this segmentation to plan grasp points, adjust orientation, and hand objects safely, ensuring dangerous regions face away from the user. Beyond drill presses and robot handoffs, this framework can be applied to multiple other applications. Together, these applications demonstrate how AI perception and real-time control can foster safer, context-aware interaction between humans and machines.