In today’s manufacturing environment, safe human–robot collaboration is of paramount importance for improving efficiency and flexibility. Targeting safety in human–robot collaboration, this chapter reports an approach for real-time collision detection and collision avoidance in an augmented virtual environment where a virtual model of a robot and real images of a human captured by depth cameras are superimposed for monitoring and collision detection. The chapter presents a distributed system which is linked to an industrial robot in a collaborative assembly cell. Four safety strategies are described: the system can alert an operator, stop a robot, move the robot away, or modify the robot’s trajectory away from an approaching operator. These strategies can be activated based on configured zones and the operator’s location with respect to the robot. The method developed is tested in realistic applications of collaboration between robots and humans in an assembly cell. Finally, the performance of the implemented approach is analysed.